language
stringclasses 6
values | original_string
stringlengths 25
887k
| text
stringlengths 25
887k
|
|---|---|---|
Python
|
def fetchChannelList(self, types=["public_channel", "private_channel", "mpim", "im"]):
"""
Method returns a list of all channel-like conversations in a workspace.
Args:
types:list<str>: Mix and match channel types by providing a comma-separated list of any combination of
public_channel, private_channel, mpim, im
"""
channels = []
logger.debug("fetch channel list for types " + str(types))
rsp = self.client.api_call(
"conversations.list",
types=", ".join(types)
)
if rsp["ok"]:
for cha in rsp["channels"]:
cha_id = cha["id"]
channels.append(cha_id)
else:
logger.error(json.dumps(rsp, indent=2))
return channels
|
def fetchChannelList(self, types=["public_channel", "private_channel", "mpim", "im"]):
"""
Method returns a list of all channel-like conversations in a workspace.
Args:
types:list<str>: Mix and match channel types by providing a comma-separated list of any combination of
public_channel, private_channel, mpim, im
"""
channels = []
logger.debug("fetch channel list for types " + str(types))
rsp = self.client.api_call(
"conversations.list",
types=", ".join(types)
)
if rsp["ok"]:
for cha in rsp["channels"]:
cha_id = cha["id"]
channels.append(cha_id)
else:
logger.error(json.dumps(rsp, indent=2))
return channels
|
Python
|
def fetchChannelHistory(self, channel):
"""
This method returns a list of all messages from the specified conversation, latest to oldest.
Args:
channel:str: channel id
Returns:
msgs:list<str>: list of messages in the form of "date user: text"
"""
hasMore = True
cur_cursor = ""
msgs = []
sleep(0.5) # dont spam the server if to much history is fechted
while hasMore:
logger.debug("fetch conversation history from " + channel)
rsp = self.client.api_call(
"conversations.history",
channel=channel,
cursor=cur_cursor
)
if rsp["ok"]:
logging.debug("has more: " + str(rsp["has_more"]))
hasMore = rsp["has_more"]
for msg in rsp["messages"]:
user = self.userIdDict[msg["user"]] # user real_name
text = msg["text"]
ts = int(msg["ts"].split('.')[0]) # unix timestamp
date = datetime.utcfromtimestamp(
ts).strftime('%Y-%m-%d %H:%M:%S')
msgs.append("{} {}: {}".format(date, user, text))
logger.debug("added {} messages to history from {}".format(
len(msgs), channel))
if hasMore:
# get next cursor
cur_cursor = rsp["response_metadata"]["next_cursor"]
else:
hasMore = False
logger.error(json.dumps(rsp, indent=2))
return msgs
|
def fetchChannelHistory(self, channel):
"""
This method returns a list of all messages from the specified conversation, latest to oldest.
Args:
channel:str: channel id
Returns:
msgs:list<str>: list of messages in the form of "date user: text"
"""
hasMore = True
cur_cursor = ""
msgs = []
sleep(0.5) # dont spam the server if to much history is fechted
while hasMore:
logger.debug("fetch conversation history from " + channel)
rsp = self.client.api_call(
"conversations.history",
channel=channel,
cursor=cur_cursor
)
if rsp["ok"]:
logging.debug("has more: " + str(rsp["has_more"]))
hasMore = rsp["has_more"]
for msg in rsp["messages"]:
user = self.userIdDict[msg["user"]] # user real_name
text = msg["text"]
ts = int(msg["ts"].split('.')[0]) # unix timestamp
date = datetime.utcfromtimestamp(
ts).strftime('%Y-%m-%d %H:%M:%S')
msgs.append("{} {}: {}".format(date, user, text))
logger.debug("added {} messages to history from {}".format(
len(msgs), channel))
if hasMore:
# get next cursor
cur_cursor = rsp["response_metadata"]["next_cursor"]
else:
hasMore = False
logger.error(json.dumps(rsp, indent=2))
return msgs
|
Python
|
def enter_rtm_loop(self, retry=1):
"""
Starts the real time messaging loop
"""
try:
if self.client.rtm_connect(with_team_state=False):
logger.info("Connected to rtm api...")
online = True
while online:
event = self.client.rtm_read()
self._parse_rtm_event(event)
sleep(1)
else:
logger.error("Connection Failed")
except TimeoutError:
logger.error("Connection timeout!")
if retry > 0:
logger.info("try to reconnect: " + str(retry))
self.enter_rtm_loop(retry=(retry - 1))
|
def enter_rtm_loop(self, retry=1):
"""
Starts the real time messaging loop
"""
try:
if self.client.rtm_connect(with_team_state=False):
logger.info("Connected to rtm api...")
online = True
while online:
event = self.client.rtm_read()
self._parse_rtm_event(event)
sleep(1)
else:
logger.error("Connection Failed")
except TimeoutError:
logger.error("Connection timeout!")
if retry > 0:
logger.info("try to reconnect: " + str(retry))
self.enter_rtm_loop(retry=(retry - 1))
|
Python
|
def _parse_rtm_event(self, event):
"""
Try to parse an JSON respons and handle it.
List of possible events and respons format under https://api.slack.com/rtm
Args:
event:json: JSON respons from the rtm websocket
"""
if len(event) == 0:
return # got nothing, pass on
rsp = event[0] # rtm event comes as an list with one or none element
try:
if rsp["type"] == "message": # got a message
if "subtype" in rsp: # has a subtype
if rsp["subtype"] == "message_deleted": # message deleted
msg = rsp["previous_message"]["text"]
logger.info("\"{}\" got deleted!".format(msg))
elif rsp["subtype"] == "message_changed": # message changed
old = rsp["previous_message"]["text"]
new = rsp["message"]["text"]
logger.info(
"\"{}\" got changed to \"{}\"".format(old, new))
else:
# unexpected rsp
logger.warning(json.dumps(event, indent=2))
else: # regular message
msg = rsp["text"]
userId = rsp["user"]
logger.info("msg: \"{}\" from \"{}\"".format(
msg, self.userIdDict[userId]))
if msg.startswith("."): # msg is a command
self._parse_command(msg, userId)
elif rsp["type"] == "hello": # server hello
logger.debug("got hello from server")
elif rsp["type"] == "user_typing": # user typing
logger.info("{} is typing".format(
self.userIdDict[rsp["user"]]))
elif rsp["type"] == "user_change": # user changed
name = rsp["user"]["real_name"]
status = rsp["user"]["profile"]["status_text"]
# in the assumption that this user entered or left a call
logger.info("{} changed. Status: {}".format(name, status))
logger.debug(json.dumps(event, indent=2))
elif rsp["type"] == "desktop_notification": # notification
logger.info("desktop_notification")
else:
logger.warning(json.dumps(event, indent=2)) # unexpected rsp
except KeyError as ke:
logger.error("KeyError: " + str(ke))
logger.error(json.dumps(event, indent=2))
|
def _parse_rtm_event(self, event):
"""
Try to parse an JSON respons and handle it.
List of possible events and respons format under https://api.slack.com/rtm
Args:
event:json: JSON respons from the rtm websocket
"""
if len(event) == 0:
return # got nothing, pass on
rsp = event[0] # rtm event comes as an list with one or none element
try:
if rsp["type"] == "message": # got a message
if "subtype" in rsp: # has a subtype
if rsp["subtype"] == "message_deleted": # message deleted
msg = rsp["previous_message"]["text"]
logger.info("\"{}\" got deleted!".format(msg))
elif rsp["subtype"] == "message_changed": # message changed
old = rsp["previous_message"]["text"]
new = rsp["message"]["text"]
logger.info(
"\"{}\" got changed to \"{}\"".format(old, new))
else:
# unexpected rsp
logger.warning(json.dumps(event, indent=2))
else: # regular message
msg = rsp["text"]
userId = rsp["user"]
logger.info("msg: \"{}\" from \"{}\"".format(
msg, self.userIdDict[userId]))
if msg.startswith("."): # msg is a command
self._parse_command(msg, userId)
elif rsp["type"] == "hello": # server hello
logger.debug("got hello from server")
elif rsp["type"] == "user_typing": # user typing
logger.info("{} is typing".format(
self.userIdDict[rsp["user"]]))
elif rsp["type"] == "user_change": # user changed
name = rsp["user"]["real_name"]
status = rsp["user"]["profile"]["status_text"]
# in the assumption that this user entered or left a call
logger.info("{} changed. Status: {}".format(name, status))
logger.debug(json.dumps(event, indent=2))
elif rsp["type"] == "desktop_notification": # notification
logger.info("desktop_notification")
else:
logger.warning(json.dumps(event, indent=2)) # unexpected rsp
except KeyError as ke:
logger.error("KeyError: " + str(ke))
logger.error(json.dumps(event, indent=2))
|
Python
|
def _system_call(self, cmd, timeout=5):
"""
Calls a given command on the system. DANGEROUS!
Don't let user input stuff to run arbitrarily commands.
Args:
cmd:list<str>: First element of the list is the command, following are arguments.
timeout:int: Seconds before the execution is aborted.
Return:
(out, err): touple of stdout and stderr from the command
"""
proc = subprocess.Popen(cmd,
stdout=subprocess.PIPE,
stderr=subprocess.PIPE)
return proc.communicate(timeout=timeout)
|
def _system_call(self, cmd, timeout=5):
"""
Calls a given command on the system. DANGEROUS!
Don't let user input stuff to run arbitrarily commands.
Args:
cmd:list<str>: First element of the list is the command, following are arguments.
timeout:int: Seconds before the execution is aborted.
Return:
(out, err): touple of stdout and stderr from the command
"""
proc = subprocess.Popen(cmd,
stdout=subprocess.PIPE,
stderr=subprocess.PIPE)
return proc.communicate(timeout=timeout)
|
Python
|
def handle_pull_request(job: PullRequestJob):
"""Analyse and handle a pull request that has just been updated."""
if job.pull_request.author == job.settings.robot:
return handle_parent_pull_request(job, job.pull_request)
try:
_handle_pull_request(job)
except messages.TemplateException as err:
send_comment(job.settings, job.pull_request, err)
raise
|
def handle_pull_request(job: PullRequestJob):
"""Analyse and handle a pull request that has just been updated."""
if job.pull_request.author == job.settings.robot:
return handle_parent_pull_request(job, job.pull_request)
try:
_handle_pull_request(job)
except messages.TemplateException as err:
send_comment(job.settings, job.pull_request, err)
raise
|
Python
|
def handle_commit(job: CommitJob):
"""Handle a job triggered by an updated build status."""
candidates = [
branch_factory(job.git.repo, b)
for b in job.git.repo.get_branches_from_commit(job.commit)
]
if not candidates:
raise messages.NothingToDo(
'Could not find any branch for commit {}' .format(job.commit)
)
if job.settings.use_queue:
if any(isinstance(b, QueueBranch) for b in candidates):
return queueing.handle_merge_queues(QueuesJob(bert_e=job.bert_e))
def get_parent_branch(branch):
if isinstance(branch, IntegrationBranch):
return branch.feature_branch
else:
return branch.name
candidates = list(map(get_parent_branch, candidates))
prs = list(
job.project_repo.get_pull_requests(src_branch=candidates)
)
if not prs:
raise messages.NothingToDo(
'Could not find the main pull request for commit {}' .format(
job.commit)
)
pr = min(prs, key=lambda pr: pr.id)
return handle_pull_request(
PullRequestJob(
bert_e=job.bert_e,
pull_request=job.project_repo.get_pull_request(int(pr.id))
)
)
|
def handle_commit(job: CommitJob):
"""Handle a job triggered by an updated build status."""
candidates = [
branch_factory(job.git.repo, b)
for b in job.git.repo.get_branches_from_commit(job.commit)
]
if not candidates:
raise messages.NothingToDo(
'Could not find any branch for commit {}' .format(job.commit)
)
if job.settings.use_queue:
if any(isinstance(b, QueueBranch) for b in candidates):
return queueing.handle_merge_queues(QueuesJob(bert_e=job.bert_e))
def get_parent_branch(branch):
if isinstance(branch, IntegrationBranch):
return branch.feature_branch
else:
return branch.name
candidates = list(map(get_parent_branch, candidates))
prs = list(
job.project_repo.get_pull_requests(src_branch=candidates)
)
if not prs:
raise messages.NothingToDo(
'Could not find the main pull request for commit {}' .format(
job.commit)
)
pr = min(prs, key=lambda pr: pr.id)
return handle_pull_request(
PullRequestJob(
bert_e=job.bert_e,
pull_request=job.project_repo.get_pull_request(int(pr.id))
)
)
|
Python
|
def handle_parent_pull_request(job, child_pr, is_child=True):
"""Handle the parent of an integration pull request."""
if is_child:
ids = re.findall(r'\d+', child_pr.description)
if not ids:
raise messages.ParentPullRequestNotFound(child_pr.id)
parent_id, *_ = ids
else:
parent_id = child_pr.id
return handle_pull_request(
PullRequestJob(
bert_e=job.bert_e,
pull_request=job.project_repo.get_pull_request(int(parent_id))
)
)
|
def handle_parent_pull_request(job, child_pr, is_child=True):
"""Handle the parent of an integration pull request."""
if is_child:
ids = re.findall(r'\d+', child_pr.description)
if not ids:
raise messages.ParentPullRequestNotFound(child_pr.id)
parent_id, *_ = ids
else:
parent_id = child_pr.id
return handle_pull_request(
PullRequestJob(
bert_e=job.bert_e,
pull_request=job.project_repo.get_pull_request(int(parent_id))
)
)
|
Python
|
def early_checks(job):
"""Early checks to filter out pull requests where no action is needed."""
status = job.pull_request.status
if status not in ('OPEN', 'DECLINED'):
raise messages.NothingToDo("The pull request is '{}'".format(status))
src, dst = job.pull_request.src_branch, job.pull_request.dst_branch
if not is_cascade_producer(src) or not is_cascade_consumer(dst):
raise messages.NotMyJob(src, dst)
if not job.git.repo.remote_branch_exists(dst):
raise messages.WrongDestination(dst_branch=dst,
active_options=job.active_options)
|
def early_checks(job):
"""Early checks to filter out pull requests where no action is needed."""
status = job.pull_request.status
if status not in ('OPEN', 'DECLINED'):
raise messages.NothingToDo("The pull request is '{}'".format(status))
src, dst = job.pull_request.src_branch, job.pull_request.dst_branch
if not is_cascade_producer(src) or not is_cascade_consumer(dst):
raise messages.NotMyJob(src, dst)
if not job.git.repo.remote_branch_exists(dst):
raise messages.WrongDestination(dst_branch=dst,
active_options=job.active_options)
|
Python
|
def send_greetings(job):
"""Send welcome message to the pull request's author and set default tasks.
"""
username = job.settings.robot
if find_comment(job.pull_request, username=username):
return
tasks = list(reversed(job.settings.tasks))
comment = send_comment(
job.settings, job.pull_request, messages.InitMessage(
bert_e=username, author=job.pull_request.author_display_name,
status={}, active_options=job.active_options, tasks=tasks,
frontend_url=job.bert_e.settings.frontend_url
)
)
for task in tasks:
create_task(job.settings, task, comment)
|
def send_greetings(job):
"""Send welcome message to the pull request's author and set default tasks.
"""
username = job.settings.robot
if find_comment(job.pull_request, username=username):
return
tasks = list(reversed(job.settings.tasks))
comment = send_comment(
job.settings, job.pull_request, messages.InitMessage(
bert_e=username, author=job.pull_request.author_display_name,
status={}, active_options=job.active_options, tasks=tasks,
frontend_url=job.bert_e.settings.frontend_url
)
)
for task in tasks:
create_task(job.settings, task, comment)
|
Python
|
def handle_comments(job):
"""Handle options and commands in the pull request's comments.
Raises:
UnknownCommand: if an unrecognized command is sent to BertE.
NotEnoughCredentials: if the author of a message is trying to set an
option or call a command he is not allowed to.
"""
reactor = Reactor()
admins = job.settings.admins
pr_author = job.pull_request.author
reactor.init_settings(job)
prefix = '@{}'.format(job.settings.robot)
LOG.debug('looking for prefix: %s', prefix)
# Handle options
# Look for options in all of the pull request's comments.
for comment in job.pull_request.comments:
author = comment.author
privileged = author in admins and author != pr_author
authored = author == pr_author
text = comment.text
try:
reactor.handle_options(job, text, prefix, privileged, authored)
except NotFound as err:
raise messages.UnknownCommand(
active_options=job.active_options, command=err.keyword,
author=author, comment=text
) from err
except NotPrivileged as err:
raise messages.NotEnoughCredentials(
active_options=job.active_options, command=err.keyword,
author=author, self_pr=(author == pr_author), comment=text
) from err
except NotAuthored as err:
raise messages.NotAuthor(
active_options=job.active_options, command=err.keyword,
author=author, pr_author=pr_author, authored=authored
) from err
except TypeError as err:
raise messages.IncorrectCommandSyntax(
extra_message=str(err), active_options=job.active_options
) from err
# Handle commands
# Look for commands in comments posted after BertE's last message.
for comment in reversed(job.pull_request.comments):
author = comment.author
if author == job.settings.robot:
return
privileged = author in admins and author != pr_author
text = comment.text
try:
reactor.handle_commands(job, text, prefix, privileged)
except NotFound as err:
raise messages.UnknownCommand(
active_options=job.active_options, command=err.keyword,
author=author, comment=text
) from err
except NotPrivileged as err:
raise messages.NotEnoughCredentials(
active_options=job.active_options, command=err.keyword,
author=author, self_pr=(author == pr_author), comment=text
) from err
|
def handle_comments(job):
"""Handle options and commands in the pull request's comments.
Raises:
UnknownCommand: if an unrecognized command is sent to BertE.
NotEnoughCredentials: if the author of a message is trying to set an
option or call a command he is not allowed to.
"""
reactor = Reactor()
admins = job.settings.admins
pr_author = job.pull_request.author
reactor.init_settings(job)
prefix = '@{}'.format(job.settings.robot)
LOG.debug('looking for prefix: %s', prefix)
# Handle options
# Look for options in all of the pull request's comments.
for comment in job.pull_request.comments:
author = comment.author
privileged = author in admins and author != pr_author
authored = author == pr_author
text = comment.text
try:
reactor.handle_options(job, text, prefix, privileged, authored)
except NotFound as err:
raise messages.UnknownCommand(
active_options=job.active_options, command=err.keyword,
author=author, comment=text
) from err
except NotPrivileged as err:
raise messages.NotEnoughCredentials(
active_options=job.active_options, command=err.keyword,
author=author, self_pr=(author == pr_author), comment=text
) from err
except NotAuthored as err:
raise messages.NotAuthor(
active_options=job.active_options, command=err.keyword,
author=author, pr_author=pr_author, authored=authored
) from err
except TypeError as err:
raise messages.IncorrectCommandSyntax(
extra_message=str(err), active_options=job.active_options
) from err
# Handle commands
# Look for commands in comments posted after BertE's last message.
for comment in reversed(job.pull_request.comments):
author = comment.author
if author == job.settings.robot:
return
privileged = author in admins and author != pr_author
text = comment.text
try:
reactor.handle_commands(job, text, prefix, privileged)
except NotFound as err:
raise messages.UnknownCommand(
active_options=job.active_options, command=err.keyword,
author=author, comment=text
) from err
except NotPrivileged as err:
raise messages.NotEnoughCredentials(
active_options=job.active_options, command=err.keyword,
author=author, self_pr=(author == pr_author), comment=text
) from err
|
Python
|
def check_commit_diff(job):
"""Check for divergence between a PR's source and destination branches.
raises:
SourceBranchTooOld: if the branches have diverged.
"""
threshold = job.settings.max_commit_diff
LOG.debug('max_commit_diff: %d', job.settings.max_commit_diff)
if threshold < 1:
# Feature is deactivated (default)
return
commits = list(job.git.dst_branch.get_commit_diff(job.git.src_branch))
LOG.debug('commit_diff: %d', len(commits))
if len(commits) > threshold:
raise messages.SourceBranchTooOld(
src_branch=job.git.src_branch.name,
dst_branch=job.git.dst_branch.name,
threshold=threshold,
active_options=job.active_options
)
|
def check_commit_diff(job):
"""Check for divergence between a PR's source and destination branches.
raises:
SourceBranchTooOld: if the branches have diverged.
"""
threshold = job.settings.max_commit_diff
LOG.debug('max_commit_diff: %d', job.settings.max_commit_diff)
if threshold < 1:
# Feature is deactivated (default)
return
commits = list(job.git.dst_branch.get_commit_diff(job.git.src_branch))
LOG.debug('commit_diff: %d', len(commits))
if len(commits) > threshold:
raise messages.SourceBranchTooOld(
src_branch=job.git.src_branch.name,
dst_branch=job.git.dst_branch.name,
threshold=threshold,
active_options=job.active_options
)
|
Python
|
def check_branch_compatibility(job):
"""Check that the pull request's source and destination branches are
compatible with one another.
For example, check that the user is not trying to merge a new feature
into any older development/* branch.
Raises:
IncompatibleSourceBranchPrefix: if the prefix of the source branch
is incorrect.
"""
if bypass_incompatible_branch(job):
return
src_branch = job.git.src_branch
for dst_branch in job.git.cascade.dst_branches:
if src_branch.prefix not in dst_branch.allow_prefixes:
raise messages.IncompatibleSourceBranchPrefix(
source=src_branch,
destination=job.git.dst_branch,
active_options=job.active_options
)
|
def check_branch_compatibility(job):
"""Check that the pull request's source and destination branches are
compatible with one another.
For example, check that the user is not trying to merge a new feature
into any older development/* branch.
Raises:
IncompatibleSourceBranchPrefix: if the prefix of the source branch
is incorrect.
"""
if bypass_incompatible_branch(job):
return
src_branch = job.git.src_branch
for dst_branch in job.git.cascade.dst_branches:
if src_branch.prefix not in dst_branch.allow_prefixes:
raise messages.IncompatibleSourceBranchPrefix(
source=src_branch,
destination=job.git.dst_branch,
active_options=job.active_options
)
|
Python
|
def check_dependencies(job):
"""Check the pull request's dependencies, if any.
Raises:
AfterPullRequest: if the current pull request depends on other open
pull requests to be merged.
NothingToDo: if the wait option is set then nothing will be checked.
"""
if job.settings.wait:
raise messages.NothingToDo('wait option is set')
after_prs = job.settings.after_pull_request
if not after_prs:
return
prs = []
for pr_id in after_prs:
try:
prs.append(job.project_repo.get_pull_request(int(pr_id)))
except Exception as err:
raise messages.IncorrectPullRequestNumber(
pr_id=pr_id, active_options=job.active_options
) from err
opened = [p for p in prs if p.status == 'OPEN']
merged = [p for p in prs if p.status == 'MERGED']
declined = [p for p in prs if p.status == 'DECLINED']
if len(after_prs) != len(merged):
raise messages.AfterPullRequest(
opened_prs=opened, declined_prs=declined,
active_options=job.active_options
)
|
def check_dependencies(job):
"""Check the pull request's dependencies, if any.
Raises:
AfterPullRequest: if the current pull request depends on other open
pull requests to be merged.
NothingToDo: if the wait option is set then nothing will be checked.
"""
if job.settings.wait:
raise messages.NothingToDo('wait option is set')
after_prs = job.settings.after_pull_request
if not after_prs:
return
prs = []
for pr_id in after_prs:
try:
prs.append(job.project_repo.get_pull_request(int(pr_id)))
except Exception as err:
raise messages.IncorrectPullRequestNumber(
pr_id=pr_id, active_options=job.active_options
) from err
opened = [p for p in prs if p.status == 'OPEN']
merged = [p for p in prs if p.status == 'MERGED']
declined = [p for p in prs if p.status == 'DECLINED']
if len(after_prs) != len(merged):
raise messages.AfterPullRequest(
opened_prs=opened, declined_prs=declined,
active_options=job.active_options
)
|
Python
|
def handle_declined_pull_request(job):
"""The pull request was declined.
Decline integration pull requests and cleanup integration branches.
Raises:
PullRequestDeclined: if some cleanup was done.
NothingToDo: if everything was already clean.
"""
build_branch_cascade(job)
changed = False
src_branch = job.pull_request.src_branch
dst_branches = job.git.cascade.dst_branches
wbranch_names = ['w/{}/{}'.format(b.version, src_branch)
for b in dst_branches]
open_prs = list(
job.project_repo.get_pull_requests(src_branch=wbranch_names)
)
for name, dst_branch in zip(wbranch_names, dst_branches):
for pr in open_prs:
if (pr.status == 'OPEN' and
pr.src_branch == name and
pr.dst_branch == dst_branch.name):
pr.decline()
changed = True
break
wbranch = branch_factory(job.git.repo, name)
wbranch.src_branch = src_branch
wbranch.dst_branch = dst_branch
if wbranch.exists():
wbranch.remove()
changed = True
if changed:
push(job.git.repo, prune=True)
raise messages.PullRequestDeclined()
else:
raise messages.NothingToDo()
|
def handle_declined_pull_request(job):
"""The pull request was declined.
Decline integration pull requests and cleanup integration branches.
Raises:
PullRequestDeclined: if some cleanup was done.
NothingToDo: if everything was already clean.
"""
build_branch_cascade(job)
changed = False
src_branch = job.pull_request.src_branch
dst_branches = job.git.cascade.dst_branches
wbranch_names = ['w/{}/{}'.format(b.version, src_branch)
for b in dst_branches]
open_prs = list(
job.project_repo.get_pull_requests(src_branch=wbranch_names)
)
for name, dst_branch in zip(wbranch_names, dst_branches):
for pr in open_prs:
if (pr.status == 'OPEN' and
pr.src_branch == name and
pr.dst_branch == dst_branch.name):
pr.decline()
changed = True
break
wbranch = branch_factory(job.git.repo, name)
wbranch.src_branch = src_branch
wbranch.dst_branch = dst_branch
if wbranch.exists():
wbranch.remove()
changed = True
if changed:
push(job.git.repo, prune=True)
raise messages.PullRequestDeclined()
else:
raise messages.NothingToDo()
|
Python
|
def check_in_sync(job, wbranches) -> bool:
"""Validate that each integration branch contains the last commit from its
predecessor.
Returns:
True: if integration branches are in sync.
False: otherwise.
"""
prev = job.git.src_branch
for branch in wbranches:
if not branch.includes_commit(prev.get_latest_commit()):
return False
prev = branch
return True
|
def check_in_sync(job, wbranches) -> bool:
"""Validate that each integration branch contains the last commit from its
predecessor.
Returns:
True: if integration branches are in sync.
False: otherwise.
"""
prev = job.git.src_branch
for branch in wbranches:
if not branch.includes_commit(prev.get_latest_commit()):
return False
prev = branch
return True
|
Python
|
def check_pull_request_skew(job, wbranches, child_prs):
"""Check potential skew between local commit and commit in PR.
Three cases are possible:
- the local commit and the commit we obtained in the PR
object are identical; nothing to do.
- the local commit, that has just been pushed by Bert-E,
does not reflect yet in the PR object we obtained from
bitbucket (the cache mechanism from BB mean the PR is still
pointing to a previous commit); the solution is to update
the PR object with the latest commit we know of.
- the local commit is outdated, someone else has pushed new
commits on the integration branch, and it reflects in the PR
object; in this case we abort the process, Bert-E will be
called again on the new commits.
Raises:
PullRequestSkewDetected: if a skew is detected.
"""
for branch, pull_request in zip(wbranches, child_prs):
branch_sha1 = branch.get_latest_commit()
pr_sha1 = pull_request.src_commit # 12 hex hash
if branch_sha1.startswith(pr_sha1):
continue
if branch.includes_commit(pr_sha1):
LOG.warning('Skew detected (expected commit: %s, '
'got PR commit: %s).', branch_sha1, pr_sha1)
LOG.warning('Updating the integration PR locally.')
pull_request.src_commit = branch_sha1
continue
raise messages.PullRequestSkewDetected(pull_request.id, branch_sha1,
pr_sha1)
|
def check_pull_request_skew(job, wbranches, child_prs):
"""Check potential skew between local commit and commit in PR.
Three cases are possible:
- the local commit and the commit we obtained in the PR
object are identical; nothing to do.
- the local commit, that has just been pushed by Bert-E,
does not reflect yet in the PR object we obtained from
bitbucket (the cache mechanism from BB mean the PR is still
pointing to a previous commit); the solution is to update
the PR object with the latest commit we know of.
- the local commit is outdated, someone else has pushed new
commits on the integration branch, and it reflects in the PR
object; in this case we abort the process, Bert-E will be
called again on the new commits.
Raises:
PullRequestSkewDetected: if a skew is detected.
"""
for branch, pull_request in zip(wbranches, child_prs):
branch_sha1 = branch.get_latest_commit()
pr_sha1 = pull_request.src_commit # 12 hex hash
if branch_sha1.startswith(pr_sha1):
continue
if branch.includes_commit(pr_sha1):
LOG.warning('Skew detected (expected commit: %s, '
'got PR commit: %s).', branch_sha1, pr_sha1)
LOG.warning('Updating the integration PR locally.')
pull_request.src_commit = branch_sha1
continue
raise messages.PullRequestSkewDetected(pull_request.id, branch_sha1,
pr_sha1)
|
Python
|
def check_approvals(job):
"""Check approval of a pull request by author, peers, and leaders.
Raises:
- ApprovalRequired
"""
required_peer_approvals = job.settings.required_peer_approvals
current_peer_approvals = 0
if bypass_peer_approval(job):
current_peer_approvals = required_peer_approvals
required_leader_approvals = job.settings.required_leader_approvals
current_leader_approvals = 0
if bypass_leader_approval(job):
current_leader_approvals = required_leader_approvals
approved_by_author = (
not job.settings.need_author_approval or
bypass_author_approval(job) or
job.settings.approve
)
requires_unanimity = job.settings.unanimity
is_unanimous = True
if (approved_by_author and
(current_peer_approvals >= required_peer_approvals) and
(current_leader_approvals >= required_leader_approvals) and
not requires_unanimity):
return
# NB: when author hasn't approved the PR, author isn't listed in
# 'participants'
username = job.settings.robot
participants = set(job.pull_request.get_participants())
approvals = set(job.pull_request.get_approvals())
if job.settings.approve:
approvals.add(job.pull_request.author)
# Exclude Bert-E from consideration
participants -= {username}
leaders = set(job.settings.project_leaders)
is_unanimous = approvals - {username} == participants
approved_by_author |= job.pull_request.author in approvals
current_leader_approvals += len(approvals.intersection(leaders))
if (job.pull_request.author in leaders and
job.pull_request.author not in approvals):
# if a project leader creates a PR and has not approved it
# (which is not possible on Github for example), always count
# one additional mandatory approval
current_leader_approvals += 1
missing_leader_approvals = (
required_leader_approvals - current_leader_approvals)
peer_approvals = approvals - {job.pull_request.author}
current_peer_approvals += len(peer_approvals)
missing_peer_approvals = (
required_peer_approvals - current_peer_approvals)
change_requests = set(job.pull_request.get_change_requests())
LOG.info('approvals: %s' % locals())
if not approved_by_author or \
missing_leader_approvals > 0 or \
missing_peer_approvals > 0 or \
(requires_unanimity and not is_unanimous) or \
len(change_requests) > 0:
raise messages.ApprovalRequired(
pr=job.pull_request,
required_leader_approvals=required_leader_approvals,
leaders=list(leaders),
required_peer_approvals=required_peer_approvals,
requires_unanimity=requires_unanimity,
requires_author_approval=job.settings.need_author_approval,
pr_author_options=job.settings.pr_author_options,
active_options=job.active_options,
change_requesters=list(change_requests)
)
|
def check_approvals(job):
"""Check approval of a pull request by author, peers, and leaders.
Raises:
- ApprovalRequired
"""
required_peer_approvals = job.settings.required_peer_approvals
current_peer_approvals = 0
if bypass_peer_approval(job):
current_peer_approvals = required_peer_approvals
required_leader_approvals = job.settings.required_leader_approvals
current_leader_approvals = 0
if bypass_leader_approval(job):
current_leader_approvals = required_leader_approvals
approved_by_author = (
not job.settings.need_author_approval or
bypass_author_approval(job) or
job.settings.approve
)
requires_unanimity = job.settings.unanimity
is_unanimous = True
if (approved_by_author and
(current_peer_approvals >= required_peer_approvals) and
(current_leader_approvals >= required_leader_approvals) and
not requires_unanimity):
return
# NB: when author hasn't approved the PR, author isn't listed in
# 'participants'
username = job.settings.robot
participants = set(job.pull_request.get_participants())
approvals = set(job.pull_request.get_approvals())
if job.settings.approve:
approvals.add(job.pull_request.author)
# Exclude Bert-E from consideration
participants -= {username}
leaders = set(job.settings.project_leaders)
is_unanimous = approvals - {username} == participants
approved_by_author |= job.pull_request.author in approvals
current_leader_approvals += len(approvals.intersection(leaders))
if (job.pull_request.author in leaders and
job.pull_request.author not in approvals):
# if a project leader creates a PR and has not approved it
# (which is not possible on Github for example), always count
# one additional mandatory approval
current_leader_approvals += 1
missing_leader_approvals = (
required_leader_approvals - current_leader_approvals)
peer_approvals = approvals - {job.pull_request.author}
current_peer_approvals += len(peer_approvals)
missing_peer_approvals = (
required_peer_approvals - current_peer_approvals)
change_requests = set(job.pull_request.get_change_requests())
LOG.info('approvals: %s' % locals())
if not approved_by_author or \
missing_leader_approvals > 0 or \
missing_peer_approvals > 0 or \
(requires_unanimity and not is_unanimous) or \
len(change_requests) > 0:
raise messages.ApprovalRequired(
pr=job.pull_request,
required_leader_approvals=required_leader_approvals,
leaders=list(leaders),
required_peer_approvals=required_peer_approvals,
requires_unanimity=requires_unanimity,
requires_author_approval=job.settings.need_author_approval,
pr_author_options=job.settings.pr_author_options,
active_options=job.active_options,
change_requesters=list(change_requests)
)
|
Python
|
def check_build_status(job, wbranches):
"""Check the build statuses of the integration pull requests.
Raises:
BuildFailed: if a build failed or was stopped.
BuildNotStarted: if a build hasn't started yet.
BuildInProgress: if a build is still in progress.
"""
if bypass_build_status(job):
return
key = job.settings.build_key
if not key:
return
ordered_state = {
status: idx for idx, status in enumerate(
('SUCCESSFUL', 'INPROGRESS', 'NOTSTARTED', 'STOPPED', 'FAILED'))
}
def status(branch):
return job.project_repo.get_build_status(
branch.get_latest_commit(), key)
statuses = {b.name: status(b) for b in wbranches}
worst = max(wbranches, key=lambda b: ordered_state[statuses[b.name]])
worst_status = statuses[worst.name]
if worst_status in ('FAILED', 'STOPPED'):
raise messages.BuildFailed(
active_options=job.active_options,
branch=worst.name,
build_url=job.project_repo.get_build_url(
worst.get_latest_commit,
key),
commit_url=job.project_repo.get_commit_url(
worst.get_latest_commit()),
)
elif worst_status == 'NOTSTARTED':
raise messages.BuildNotStarted()
elif worst_status == 'INPROGRESS':
raise messages.BuildInProgress()
assert worst_status == 'SUCCESSFUL'
|
def check_build_status(job, wbranches):
"""Check the build statuses of the integration pull requests.
Raises:
BuildFailed: if a build failed or was stopped.
BuildNotStarted: if a build hasn't started yet.
BuildInProgress: if a build is still in progress.
"""
if bypass_build_status(job):
return
key = job.settings.build_key
if not key:
return
ordered_state = {
status: idx for idx, status in enumerate(
('SUCCESSFUL', 'INPROGRESS', 'NOTSTARTED', 'STOPPED', 'FAILED'))
}
def status(branch):
return job.project_repo.get_build_status(
branch.get_latest_commit(), key)
statuses = {b.name: status(b) for b in wbranches}
worst = max(wbranches, key=lambda b: ordered_state[statuses[b.name]])
worst_status = statuses[worst.name]
if worst_status in ('FAILED', 'STOPPED'):
raise messages.BuildFailed(
active_options=job.active_options,
branch=worst.name,
build_url=job.project_repo.get_build_url(
worst.get_latest_commit,
key),
commit_url=job.project_repo.get_commit_url(
worst.get_latest_commit()),
)
elif worst_status == 'NOTSTARTED':
raise messages.BuildNotStarted()
elif worst_status == 'INPROGRESS':
raise messages.BuildInProgress()
assert worst_status == 'SUCCESSFUL'
|
Python
|
def _patch_url(self, url):
"""Patch URLs if it is relative to the API root.
Returns: an absolute url corresponding to client.base_url / url
"""
if not url.startswith('http'):
url = '/'.join((self.base_url, url.lstrip('/')))
return url
|
def _patch_url(self, url):
"""Patch URLs if it is relative to the API root.
Returns: an absolute url corresponding to client.base_url / url
"""
if not url.startswith('http'):
url = '/'.join((self.base_url, url.lstrip('/')))
return url
|
Python
|
def _cache_value(self, method, url, params, res):
"""Put a request result in the query cache.
If the response's headers contain an ETag or a Last-Modified field,
the response can be used in subsequent calls to avoid hitting github's
rate limit.
Args:
- method (str): request method (e.g. GET)
- url (str): request url
- params (dict): request parameter dict as per requests library's
params argument
- res (requests.Response): the request's response
Returns:
The response that was put in cache.
"""
key = self._mk_key(url, params)
headers = res.headers
etag = headers.get('ETag', None)
date = headers.get('Last-Modified', None)
if etag or date:
self.query_cache[method].set(key, CacheEntry(res, etag, date))
return res
|
def _cache_value(self, method, url, params, res):
"""Put a request result in the query cache.
If the response's headers contain an ETag or a Last-Modified field,
the response can be used in subsequent calls to avoid hitting github's
rate limit.
Args:
- method (str): request method (e.g. GET)
- url (str): request url
- params (dict): request parameter dict as per requests library's
params argument
- res (requests.Response): the request's response
Returns:
The response that was put in cache.
"""
key = self._mk_key(url, params)
headers = res.headers
etag = headers.get('ETag', None)
date = headers.get('Last-Modified', None)
if etag or date:
self.query_cache[method].set(key, CacheEntry(res, etag, date))
return res
|
Python
|
def _get_cached_value(self, method, url, params):
"""Get a value from the cache if any.
This method is intended to be called before performing an HTTP request,
in order to define the special headers used by GitHub's rate-limit
system.
If the request that follows returns a HTTP 304 code, this means that:
- the cached value returned by this method can be returned as a
valid result
- the request wasn't decremented from GitHub's rate limit counter
Args:
- method (str): request method (e.g. GET)
- url (str): request url
- params (dict): request parameter dict as per requests library's
params argument
Returns:
A (response, headers) tuple.
- response is the last response we've received for this request
(possibly None).
- headers is a dictionary defining 'If-None-Match' and
'If-Modified-Since' headers to add to the request.
See: the _get() method to understand how it is used.
"""
key = self._mk_key(url, params)
entry = self.query_cache[method].get(key, None)
headers = {
'If-None-Match': None,
'If-Modified-Since': None
}
if entry is None:
return None, headers
if entry.etag:
headers['If-None-Match'] = entry.etag
elif entry.date:
headers['If-Modified-Since'] = entry.date
return entry.obj, headers
|
def _get_cached_value(self, method, url, params):
"""Get a value from the cache if any.
This method is intended to be called before performing an HTTP request,
in order to define the special headers used by GitHub's rate-limit
system.
If the request that follows returns a HTTP 304 code, this means that:
- the cached value returned by this method can be returned as a
valid result
- the request wasn't decremented from GitHub's rate limit counter
Args:
- method (str): request method (e.g. GET)
- url (str): request url
- params (dict): request parameter dict as per requests library's
params argument
Returns:
A (response, headers) tuple.
- response is the last response we've received for this request
(possibly None).
- headers is a dictionary defining 'If-None-Match' and
'If-Modified-Since' headers to add to the request.
See: the _get() method to understand how it is used.
"""
key = self._mk_key(url, params)
entry = self.query_cache[method].get(key, None)
headers = {
'If-None-Match': None,
'If-Modified-Since': None
}
if entry is None:
return None, headers
if entry.etag:
headers['If-None-Match'] = entry.etag
elif entry.date:
headers['If-Modified-Since'] = entry.date
return entry.obj, headers
|
Python
|
def _get(self, url, **kwargs):
"""Perform a GET request using the rate-limit cache system.
This method is not supposed to be called by other objects. Instead, it
is wrapped by the get() and iter_get() methods.
Returns:
A requests.Response object
"""
params = kwargs.get('params', {})
url = self._patch_url(url)
res, headers = self._get_cached_value('GET', url, params)
if headers:
kwargs.setdefault('headers', {}).update(headers)
response = self.session.get(url, **kwargs)
if response.status_code == 304:
LOG.debug('Not Modified. Returning cached result')
return res
response.raise_for_status()
return self._cache_value('GET', url, params, response)
|
def _get(self, url, **kwargs):
"""Perform a GET request using the rate-limit cache system.
This method is not supposed to be called by other objects. Instead, it
is wrapped by the get() and iter_get() methods.
Returns:
A requests.Response object
"""
params = kwargs.get('params', {})
url = self._patch_url(url)
res, headers = self._get_cached_value('GET', url, params)
if headers:
kwargs.setdefault('headers', {}).update(headers)
response = self.session.get(url, **kwargs)
if response.status_code == 304:
LOG.debug('Not Modified. Returning cached result')
return res
response.raise_for_status()
return self._cache_value('GET', url, params, response)
|
Python
|
def post(self, url, data, **kwargs):
"""Perform a POST request to the github API.
Args: same as requests.post()
Returns: a deserialized json structure
Raises:
requests.HTTPError
"""
url = self._patch_url(url)
response = self.session.post(url, data=data, **kwargs)
response.raise_for_status()
return json.loads(response.text)
|
def post(self, url, data, **kwargs):
"""Perform a POST request to the github API.
Args: same as requests.post()
Returns: a deserialized json structure
Raises:
requests.HTTPError
"""
url = self._patch_url(url)
response = self.session.post(url, data=data, **kwargs)
response.raise_for_status()
return json.loads(response.text)
|
Python
|
def patch(self, url, data, **kwargs):
"""Perform a PATCH request to the github API.
Args: same as requests.patch()
Returns: a deserialized json structure
Raises:
requests.HTTPError
"""
url = self._patch_url(url)
response = self.session.post(url, data=data, **kwargs)
response.raise_for_status()
return json.loads(response.text)
|
def patch(self, url, data, **kwargs):
"""Perform a PATCH request to the github API.
Args: same as requests.patch()
Returns: a deserialized json structure
Raises:
requests.HTTPError
"""
url = self._patch_url(url)
response = self.session.post(url, data=data, **kwargs)
response.raise_for_status()
return json.loads(response.text)
|
Python
|
def delete(self, url, **kwargs):
"""Perform a DELETE request on the github API.
Args: same as requests.delete()
Raises: requests.HTTPError
"""
url = self._patch_url(url)
response = self.session.delete(url, **kwargs)
response.raise_for_status()
|
def delete(self, url, **kwargs):
"""Perform a DELETE request on the github API.
Args: same as requests.delete()
Raises: requests.HTTPError
"""
url = self._patch_url(url)
response = self.session.delete(url, **kwargs)
response.raise_for_status()
|
Python
|
def put(self, url, **kwargs):
"""Perform a PUT request to the Github API.
Args: same as requests.put()
Raises: requests.HTTPError
"""
url = self._patch_url(url)
response = self.session.put(url, **kwargs)
response.raise_for_status()
|
def put(self, url, **kwargs):
"""Perform a PUT request to the Github API.
Args: same as requests.put()
Raises: requests.HTTPError
"""
url = self._patch_url(url)
response = self.session.put(url, **kwargs)
response.raise_for_status()
|
Python
|
def iter_get(self, url, per_page=100, **kwargs):
"""Perform a paginated GET request to the Github API.
This method handles cache verfication and uses conditional requests
+ a local cache to avoid consuming API calls as counted by Github's
rate limit system.
Args:
- per_page: number of objects to get per page (max & default: 100)
- same as requests.get()
Yields: deserialized json structures
Raises: requests.HTTPError
"""
params = kwargs.setdefault('params', {})
params.setdefault('per_page', per_page)
next_page = url
while next_page:
response = self._get(next_page, **kwargs)
yield from json.loads(response.text)
next_page = None
if 'link' not in response.headers:
break
for link_rel in response.headers['link'].split(','):
link, rel = link_rel.split(';')
if 'rel="next"' in rel.strip():
next_page = link.strip(' <>')
break
# Params are already contained in the next page's url
params.clear()
|
def iter_get(self, url, per_page=100, **kwargs):
"""Perform a paginated GET request to the Github API.
This method handles cache verfication and uses conditional requests
+ a local cache to avoid consuming API calls as counted by Github's
rate limit system.
Args:
- per_page: number of objects to get per page (max & default: 100)
- same as requests.get()
Yields: deserialized json structures
Raises: requests.HTTPError
"""
params = kwargs.setdefault('params', {})
params.setdefault('per_page', per_page)
next_page = url
while next_page:
response = self._get(next_page, **kwargs)
yield from json.loads(response.text)
next_page = None
if 'link' not in response.headers:
break
for link_rel in response.headers['link'].split(','):
link, rel = link_rel.split(';')
if 'rel="next"' in rel.strip():
next_page = link.strip(' <>')
break
# Params are already contained in the next page's url
params.clear()
|
Python
|
def pull_request(self) -> PullRequest:
"""Get the PullRequest associated with this issue comment event."""
pr_dict = self.data['issue'].get('pull_request')
if pr_dict:
try:
return PullRequest.get(client=self.client, url=pr_dict['url'])
except HTTPError:
LOG.error("No pull request at url %s", pr_dict['url'])
LOG.debug("Issue #%d is not a pull request",
self.data['issue']['number'])
|
def pull_request(self) -> PullRequest:
"""Get the PullRequest associated with this issue comment event."""
pr_dict = self.data['issue'].get('pull_request')
if pr_dict:
try:
return PullRequest.get(client=self.client, url=pr_dict['url'])
except HTTPError:
LOG.error("No pull request at url %s", pr_dict['url'])
LOG.debug("Issue #%d is not a pull request",
self.data['issue']['number'])
|
Python
|
def build(self, repo):
"""Collect q branches from repository, add them to the collection."""
cmd = 'git branch -r --list origin/q/*'
for branch in repo.cmd(cmd).split('\n')[:-1]:
match_ = re.match(r'\s*origin/(?P<name>.*)', branch)
if not match_:
continue
try:
branch = branch_factory(repo, match_.group('name'))
except errors.UnrecognizedBranchPattern:
continue
self._add_branch(branch)
self.finalize()
|
def build(self, repo):
"""Collect q branches from repository, add them to the collection."""
cmd = 'git branch -r --list origin/q/*'
for branch in repo.cmd(cmd).split('\n')[:-1]:
match_ = re.match(r'\s*origin/(?P<name>.*)', branch)
if not match_:
continue
try:
branch = branch_factory(repo, match_.group('name'))
except errors.UnrecognizedBranchPattern:
continue
self._add_branch(branch)
self.finalize()
|
Python
|
def mergeable_queues(self):
"""Return a collection of queues suitable for merge.
This only works after the collection is validated.
"""
if self._mergeable_queues is None:
self._process()
return self._mergeable_queues
|
def mergeable_queues(self):
"""Return a collection of queues suitable for merge.
This only works after the collection is validated.
"""
if self._mergeable_queues is None:
self._process()
return self._mergeable_queues
|
Python
|
def mergeable_prs(self):
"""Return the list of pull requests suitable for merge.
This only works after the collection is validated.
"""
if self._mergeable_queues is None:
self._process()
return self._mergeable_prs
|
def mergeable_prs(self):
"""Return the list of pull requests suitable for merge.
This only works after the collection is validated.
"""
if self._mergeable_queues is None:
self._process()
return self._mergeable_prs
|
Python
|
def _add_branch(self, branch):
"""Add a single branch to the queue collection."""
if not isinstance(branch, (QueueBranch, QueueIntegrationBranch)):
raise errors.InvalidQueueBranch(branch)
self._validated = False
# make sure we have a local copy of the branch
# (enables get_latest_commit)
branch.checkout()
version = branch.version_t
if version not in self._queues.keys():
self._queues[version] = {
QueueBranch: None,
QueueIntegrationBranch: []
}
# Sort the top dict again
self._queues = OrderedDict(sorted(self._queues.items(),
key=cmp_to_key(compare_queues)))
if isinstance(branch, QueueBranch):
self._queues[version][QueueBranch] = branch
else:
self._queues[version][QueueIntegrationBranch].append(branch)
|
def _add_branch(self, branch):
"""Add a single branch to the queue collection."""
if not isinstance(branch, (QueueBranch, QueueIntegrationBranch)):
raise errors.InvalidQueueBranch(branch)
self._validated = False
# make sure we have a local copy of the branch
# (enables get_latest_commit)
branch.checkout()
version = branch.version_t
if version not in self._queues.keys():
self._queues[version] = {
QueueBranch: None,
QueueIntegrationBranch: []
}
# Sort the top dict again
self._queues = OrderedDict(sorted(self._queues.items(),
key=cmp_to_key(compare_queues)))
if isinstance(branch, QueueBranch):
self._queues[version][QueueBranch] = branch
else:
self._queues[version][QueueIntegrationBranch].append(branch)
|
Python
|
def _horizontal_validation(self, version):
"""Validation of the queue collection on one given version.
Called by validate().
"""
masterq = self._queues[version][QueueBranch]
# check master queue state
if not masterq:
yield errors.MasterQueueMissing(version)
else:
if not masterq.includes_commit(masterq.dst_branch):
yield errors.MasterQueueLateVsDev(masterq, masterq.dst_branch)
if not self._queues[version][QueueIntegrationBranch]:
# check master queue points to dev
if (masterq.get_latest_commit() !=
masterq.dst_branch.get_latest_commit()):
yield errors.MasterQueueNotInSync(masterq,
masterq.dst_branch)
else:
# check state of master queue wrt to greatest integration
# queue
greatest_intq = (
self._queues[version][QueueIntegrationBranch][0]
)
if (greatest_intq.get_latest_commit() !=
masterq.get_latest_commit()):
if greatest_intq.includes_commit(masterq):
yield errors.MasterQueueLateVsInt(masterq,
greatest_intq)
elif masterq.includes_commit(greatest_intq):
yield errors.MasterQueueYoungerThanInt(masterq,
greatest_intq)
else:
yield errors.MasterQueueDiverged(masterq,
greatest_intq)
# check each integration queue contains the previous one
nextq = masterq
for intq in self._queues[version][QueueIntegrationBranch]:
if not nextq.includes_commit(intq):
yield errors.QueueInclusionIssue(nextq, intq)
nextq = intq
if not nextq.includes_commit(masterq.dst_branch):
yield errors.QueueInclusionIssue(nextq, masterq.dst_branch)
|
def _horizontal_validation(self, version):
"""Validation of the queue collection on one given version.
Called by validate().
"""
masterq = self._queues[version][QueueBranch]
# check master queue state
if not masterq:
yield errors.MasterQueueMissing(version)
else:
if not masterq.includes_commit(masterq.dst_branch):
yield errors.MasterQueueLateVsDev(masterq, masterq.dst_branch)
if not self._queues[version][QueueIntegrationBranch]:
# check master queue points to dev
if (masterq.get_latest_commit() !=
masterq.dst_branch.get_latest_commit()):
yield errors.MasterQueueNotInSync(masterq,
masterq.dst_branch)
else:
# check state of master queue wrt to greatest integration
# queue
greatest_intq = (
self._queues[version][QueueIntegrationBranch][0]
)
if (greatest_intq.get_latest_commit() !=
masterq.get_latest_commit()):
if greatest_intq.includes_commit(masterq):
yield errors.MasterQueueLateVsInt(masterq,
greatest_intq)
elif masterq.includes_commit(greatest_intq):
yield errors.MasterQueueYoungerThanInt(masterq,
greatest_intq)
else:
yield errors.MasterQueueDiverged(masterq,
greatest_intq)
# check each integration queue contains the previous one
nextq = masterq
for intq in self._queues[version][QueueIntegrationBranch]:
if not nextq.includes_commit(intq):
yield errors.QueueInclusionIssue(nextq, intq)
nextq = intq
if not nextq.includes_commit(masterq.dst_branch):
yield errors.QueueInclusionIssue(nextq, masterq.dst_branch)
|
Python
|
def _vertical_validation(self, stack, versions):
"""Validation of the queue collection on one given merge path.
Called by validate().
"""
prs = self._extract_pr_ids(stack)
last_version = versions[-1]
hf_detected = False
if len(list(stack.keys())) == 1:
if len(list(stack.keys())[0]) == 4:
hf_detected = True
# check all subsequent versions have a master queue
has_queues = False
for version in versions:
if version not in stack:
if has_queues and not hf_detected:
yield errors.MasterQueueMissing(version)
continue
has_queues = True
if not stack[version][QueueBranch]:
yield errors.MasterQueueMissing(version)
# check queues are sync'ed vertically and included in each other
# (last one corresponds to same PR on all versions..., and so on)
# other way to say it: each version has all the PR_ids of the
# previous version
if last_version in stack:
while stack[last_version][QueueIntegrationBranch]:
next_vqint = stack[last_version][QueueIntegrationBranch].pop(0)
pr = next_vqint.pr_id
if pr not in prs:
# early fail
break
for version in reversed(versions[:-1]):
if version not in stack:
# supposedly finished
break
if len(version) == 4:
# skip hf from check loop
continue
if (stack[version][QueueIntegrationBranch] and
stack[version][QueueIntegrationBranch][0].pr_id ==
pr):
vqint = stack[version][QueueIntegrationBranch].pop(0)
# take this opportunity to check vertical inclusion
if not next_vqint.includes_commit(vqint):
yield errors.QueueInclusionIssue(next_vqint, vqint)
next_vqint = vqint
else:
# this pr is supposedly entirely removed from the stack
# if it comes back again, its an error
break
prs.remove(pr)
# skip hf from stack and prs before final checks
for version in versions:
if len(version) == 4:
if version not in stack:
continue
while stack[version][QueueIntegrationBranch]:
pr_id = stack[version][QueueIntegrationBranch][0].pr_id
stack[version][QueueIntegrationBranch].pop(0)
if pr_id in prs:
prs.remove(pr_id)
if prs:
# after this algorithm prs should be empty
yield errors.QueueInconsistentPullRequestsOrder()
else:
# and stack should be empty too
for version in versions:
if (version in stack and
stack[version][QueueIntegrationBranch]):
yield errors.QueueInconsistentPullRequestsOrder()
|
def _vertical_validation(self, stack, versions):
"""Validation of the queue collection on one given merge path.
Called by validate().
"""
prs = self._extract_pr_ids(stack)
last_version = versions[-1]
hf_detected = False
if len(list(stack.keys())) == 1:
if len(list(stack.keys())[0]) == 4:
hf_detected = True
# check all subsequent versions have a master queue
has_queues = False
for version in versions:
if version not in stack:
if has_queues and not hf_detected:
yield errors.MasterQueueMissing(version)
continue
has_queues = True
if not stack[version][QueueBranch]:
yield errors.MasterQueueMissing(version)
# check queues are sync'ed vertically and included in each other
# (last one corresponds to same PR on all versions..., and so on)
# other way to say it: each version has all the PR_ids of the
# previous version
if last_version in stack:
while stack[last_version][QueueIntegrationBranch]:
next_vqint = stack[last_version][QueueIntegrationBranch].pop(0)
pr = next_vqint.pr_id
if pr not in prs:
# early fail
break
for version in reversed(versions[:-1]):
if version not in stack:
# supposedly finished
break
if len(version) == 4:
# skip hf from check loop
continue
if (stack[version][QueueIntegrationBranch] and
stack[version][QueueIntegrationBranch][0].pr_id ==
pr):
vqint = stack[version][QueueIntegrationBranch].pop(0)
# take this opportunity to check vertical inclusion
if not next_vqint.includes_commit(vqint):
yield errors.QueueInclusionIssue(next_vqint, vqint)
next_vqint = vqint
else:
# this pr is supposedly entirely removed from the stack
# if it comes back again, its an error
break
prs.remove(pr)
# skip hf from stack and prs before final checks
for version in versions:
if len(version) == 4:
if version not in stack:
continue
while stack[version][QueueIntegrationBranch]:
pr_id = stack[version][QueueIntegrationBranch][0].pr_id
stack[version][QueueIntegrationBranch].pop(0)
if pr_id in prs:
prs.remove(pr_id)
if prs:
# after this algorithm prs should be empty
yield errors.QueueInconsistentPullRequestsOrder()
else:
# and stack should be empty too
for version in versions:
if (version in stack and
stack[version][QueueIntegrationBranch]):
yield errors.QueueInconsistentPullRequestsOrder()
|
Python
|
def validate(self):
"""Check the state of queues declared via add_branch.
The following checks are performed:
- horizontal checks: on a given branch version, each integration
queue must include the previous one; the master queue must
point to the last integration queue; In case there is no
integration queue (nothing queued for this version), the
master queue must point on the corresponding development
branch.
- vertical checks: across versions, on each merge path, the queues
must be in the correct order (pr1 queued first, pr2 then, ...);
when a pr is queued in a version, it must be present in all the
following integration queues; for a given pr, the queues must be
included in each other.
- completeness: in order to detect missing integration queues
(manuel delete for example), deconstruct the master queues
by reverting merge commits; each result should not differ in
content from the previous integration queue; The last diff is
checked vs the corresponding development branch. TODO
"""
errs = []
versions = self._queues.keys()
if not versions:
# no queues, cool stuff
self._validated = True
return
for version in versions:
errs.extend(self._horizontal_validation(version))
for merge_path in self.merge_paths:
versions = [branch.version_t for branch in merge_path]
stack = deepcopy(self._queues)
# remove versions not on this merge_path from consideration
for version in list(stack.keys()):
if version not in versions:
stack.pop(version)
errs.extend(self._vertical_validation(stack, versions))
if errs:
raise errors.IncoherentQueues(errs)
self._validated = True
|
def validate(self):
"""Check the state of queues declared via add_branch.
The following checks are performed:
- horizontal checks: on a given branch version, each integration
queue must include the previous one; the master queue must
point to the last integration queue; In case there is no
integration queue (nothing queued for this version), the
master queue must point on the corresponding development
branch.
- vertical checks: across versions, on each merge path, the queues
must be in the correct order (pr1 queued first, pr2 then, ...);
when a pr is queued in a version, it must be present in all the
following integration queues; for a given pr, the queues must be
included in each other.
- completeness: in order to detect missing integration queues
(manuel delete for example), deconstruct the master queues
by reverting merge commits; each result should not differ in
content from the previous integration queue; The last diff is
checked vs the corresponding development branch. TODO
"""
errs = []
versions = self._queues.keys()
if not versions:
# no queues, cool stuff
self._validated = True
return
for version in versions:
errs.extend(self._horizontal_validation(version))
for merge_path in self.merge_paths:
versions = [branch.version_t for branch in merge_path]
stack = deepcopy(self._queues)
# remove versions not on this merge_path from consideration
for version in list(stack.keys()):
if version not in versions:
stack.pop(version)
errs.extend(self._vertical_validation(stack, versions))
if errs:
raise errors.IncoherentQueues(errs)
self._validated = True
|
Python
|
def _recursive_lookup(self, queues):
"""Given a set of queues, remove all queues that can't be merged,
based on the build status obtained from the repository manager.
A pull request must be removed from the list if the build on at least
one version is FAILED, and if this failure is not recovered by
a later pull request.
Return once a mergeable set is identified or the set is empty.
"""
first_failed_pr = 0
for version in queues.keys():
qints = queues[version][QueueIntegrationBranch]
if qints:
qint = qints[0]
status = self.bbrepo.get_build_status(
qint.get_latest_commit(),
self.build_key
)
if status != 'SUCCESSFUL':
first_failed_pr = qint.pr_id
break
if first_failed_pr == 0:
# all tip queues are pass, merge as it is
return
# remove all queues that don't pass globally,
# up to the identified failed pr, and retry
for version in queues.keys():
intqs = queues[version][QueueIntegrationBranch]
if all([inq.pr_id != first_failed_pr for inq in intqs]):
# do not pop anything if failed pr is not on the current path
continue
while intqs:
intq = intqs.pop(0)
if intq.pr_id == first_failed_pr:
break
self._recursive_lookup(queues)
|
def _recursive_lookup(self, queues):
"""Given a set of queues, remove all queues that can't be merged,
based on the build status obtained from the repository manager.
A pull request must be removed from the list if the build on at least
one version is FAILED, and if this failure is not recovered by
a later pull request.
Return once a mergeable set is identified or the set is empty.
"""
first_failed_pr = 0
for version in queues.keys():
qints = queues[version][QueueIntegrationBranch]
if qints:
qint = qints[0]
status = self.bbrepo.get_build_status(
qint.get_latest_commit(),
self.build_key
)
if status != 'SUCCESSFUL':
first_failed_pr = qint.pr_id
break
if first_failed_pr == 0:
# all tip queues are pass, merge as it is
return
# remove all queues that don't pass globally,
# up to the identified failed pr, and retry
for version in queues.keys():
intqs = queues[version][QueueIntegrationBranch]
if all([inq.pr_id != first_failed_pr for inq in intqs]):
# do not pop anything if failed pr is not on the current path
continue
while intqs:
intq = intqs.pop(0)
if intq.pr_id == first_failed_pr:
break
self._recursive_lookup(queues)
|
Python
|
def _extract_pr_ids(self, queues):
"""Return list of pull requests present in a set of queues.
This is obtained by reading pr ids from the greatest
development queue branch, so assumes that this branch
contains a reference to all pull requests in queues (this
is normally the case if everything was queued by Bert-E.
Return (list):
pull request ids in provided queue set (in the order
of addition to the queue, from oldest to newest)
"""
prs_hf = []
prs = []
# identify version corresponding to last dev queue
# (i.e. ignore stab queues)
greatest_dev = None
for version in reversed(queues.keys()):
if len(version) == 2 and greatest_dev is None:
greatest_dev = version
if len(version) == 4:
# we may not catch the hf pr_id later from greatest_dev
# so insert them now
for qint in queues[version][QueueIntegrationBranch]:
if qint.pr_id not in prs_hf:
prs_hf.insert(0, qint.pr_id)
if greatest_dev:
for qint in queues[greatest_dev][QueueIntegrationBranch]:
if qint.pr_id not in prs_hf + prs:
prs.insert(0, qint.pr_id)
return prs_hf + prs
|
def _extract_pr_ids(self, queues):
"""Return list of pull requests present in a set of queues.
This is obtained by reading pr ids from the greatest
development queue branch, so assumes that this branch
contains a reference to all pull requests in queues (this
is normally the case if everything was queued by Bert-E.
Return (list):
pull request ids in provided queue set (in the order
of addition to the queue, from oldest to newest)
"""
prs_hf = []
prs = []
# identify version corresponding to last dev queue
# (i.e. ignore stab queues)
greatest_dev = None
for version in reversed(queues.keys()):
if len(version) == 2 and greatest_dev is None:
greatest_dev = version
if len(version) == 4:
# we may not catch the hf pr_id later from greatest_dev
# so insert them now
for qint in queues[version][QueueIntegrationBranch]:
if qint.pr_id not in prs_hf:
prs_hf.insert(0, qint.pr_id)
if greatest_dev:
for qint in queues[greatest_dev][QueueIntegrationBranch]:
if qint.pr_id not in prs_hf + prs:
prs.insert(0, qint.pr_id)
return prs_hf + prs
|
Python
|
def _remove_unmergeable(self, prs, queues):
"""Given a set of queues, remove all queues that are not in
the provided list of pull request ids.
"""
for version in queues.keys():
while (queues[version][QueueIntegrationBranch] and
queues[version][QueueIntegrationBranch][0].pr_id not in
prs):
queues[version][QueueIntegrationBranch].pop(0)
|
def _remove_unmergeable(self, prs, queues):
"""Given a set of queues, remove all queues that are not in
the provided list of pull request ids.
"""
for version in queues.keys():
while (queues[version][QueueIntegrationBranch] and
queues[version][QueueIntegrationBranch][0].pr_id not in
prs):
queues[version][QueueIntegrationBranch].pop(0)
|
Python
|
def _process(self):
"""Given a sorted list of queues, identify most buildable series.
We need to look at mergeable PRs from the point of view
of all the possible merge_paths individually, then merge
the results in a super-mergeable status.
Populates:
- _mergeable_queues (list): queues corresponding to the
mergeable PRs
- _mergeable_prs (list): pull requests affected by the merge
"""
if not self._validated:
raise errors.QueuesNotValidated()
mergeable_prs = self._extract_pr_ids(self._queues)
if not self.force_merge:
for merge_path in self.merge_paths:
versions = [branch.version_t for branch in merge_path]
stack = deepcopy(self._queues)
# remove versions not on this merge_path from consideration
for version in list(stack.keys()):
# exclude hf version from this pop process
if version not in versions and len(version) < 4:
stack.pop(version)
# obtain list of mergeable prs on this merge_path
self._recursive_lookup(stack)
path_mergeable_prs = self._extract_pr_ids(stack)
# smallest table is the common denominator
if len(path_mergeable_prs) < len(mergeable_prs):
mergeable_prs = path_mergeable_prs
self._mergeable_prs = mergeable_prs
mergeable_queues = deepcopy(self._queues)
self._remove_unmergeable(mergeable_prs, mergeable_queues)
self._mergeable_queues = mergeable_queues
|
def _process(self):
"""Given a sorted list of queues, identify most buildable series.
We need to look at mergeable PRs from the point of view
of all the possible merge_paths individually, then merge
the results in a super-mergeable status.
Populates:
- _mergeable_queues (list): queues corresponding to the
mergeable PRs
- _mergeable_prs (list): pull requests affected by the merge
"""
if not self._validated:
raise errors.QueuesNotValidated()
mergeable_prs = self._extract_pr_ids(self._queues)
if not self.force_merge:
for merge_path in self.merge_paths:
versions = [branch.version_t for branch in merge_path]
stack = deepcopy(self._queues)
# remove versions not on this merge_path from consideration
for version in list(stack.keys()):
# exclude hf version from this pop process
if version not in versions and len(version) < 4:
stack.pop(version)
# obtain list of mergeable prs on this merge_path
self._recursive_lookup(stack)
path_mergeable_prs = self._extract_pr_ids(stack)
# smallest table is the common denominator
if len(path_mergeable_prs) < len(mergeable_prs):
mergeable_prs = path_mergeable_prs
self._mergeable_prs = mergeable_prs
mergeable_queues = deepcopy(self._queues)
self._remove_unmergeable(mergeable_prs, mergeable_queues)
self._mergeable_queues = mergeable_queues
|
Python
|
def finalize(self):
"""Finalize the collection of queues.
Assumes _queues has been populated by calls to add_branch.
"""
# order integration queues by content
for version in self._queues.keys():
self._queues[version][
QueueIntegrationBranch].sort(reverse=True)
|
def finalize(self):
"""Finalize the collection of queues.
Assumes _queues has been populated by calls to add_branch.
"""
# order integration queues by content
for version in self._queues.keys():
self._queues[version][
QueueIntegrationBranch].sort(reverse=True)
|
Python
|
def queued_prs(self):
"""Ordered list of queued PR IDs (oldest first)."""
if not self._queues:
return []
# Find last_entry for which there is not a hf entry
last_entry = None
pr_ids = []
for key in list(reversed(self._queues.keys())):
if len(key) < 4:
last_entry = self._queues[key]
break
if last_entry is not None:
pr_ids = list(reversed([branch.pr_id for branch in
last_entry[QueueIntegrationBranch]]))
# Add hotfix PRs that are not seen from the queues top key
pr_hf_ids = []
for key in list(reversed(self._queues.keys())):
if len(key) == 4:
entry = self._queues[key]
new_pr_ids = list([branch.pr_id for branch in
entry[QueueIntegrationBranch]])
for pr_hf_id in new_pr_ids:
if pr_hf_id not in pr_hf_ids:
pr_hf_ids = [pr_hf_id] + pr_hf_ids
# Remove hotfix PRs from the first set
pr_non_hf_ids = []
for pr_id in pr_ids:
if pr_id not in pr_hf_ids:
pr_non_hf_ids = pr_non_hf_ids + [pr_id]
return pr_hf_ids + pr_non_hf_ids
|
def queued_prs(self):
"""Ordered list of queued PR IDs (oldest first)."""
if not self._queues:
return []
# Find last_entry for which there is not a hf entry
last_entry = None
pr_ids = []
for key in list(reversed(self._queues.keys())):
if len(key) < 4:
last_entry = self._queues[key]
break
if last_entry is not None:
pr_ids = list(reversed([branch.pr_id for branch in
last_entry[QueueIntegrationBranch]]))
# Add hotfix PRs that are not seen from the queues top key
pr_hf_ids = []
for key in list(reversed(self._queues.keys())):
if len(key) == 4:
entry = self._queues[key]
new_pr_ids = list([branch.pr_id for branch in
entry[QueueIntegrationBranch]])
for pr_hf_id in new_pr_ids:
if pr_hf_id not in pr_hf_ids:
pr_hf_ids = [pr_hf_id] + pr_hf_ids
# Remove hotfix PRs from the first set
pr_non_hf_ids = []
for pr_id in pr_ids:
if pr_id not in pr_hf_ids:
pr_non_hf_ids = pr_non_hf_ids + [pr_id]
return pr_hf_ids + pr_non_hf_ids
|
Python
|
def update_micro(self, tag):
"""Update development branch latest micro based on tag."""
pattern = r"^(?P<major>\d+)\.(?P<minor>\d+)\.(?P<micro>\d+)" \
r"(\.(?P<hfrev>\d+)|)$"
match = re.match(pattern, tag)
if not match:
LOG.debug("Ignore tag: %s", tag)
return
LOG.debug("Consider tag: %s", tag)
major = int(match.groupdict()['major'])
minor = int(match.groupdict()['minor'])
micro = int(match.groupdict()['micro'])
hfrev = 0 # default hfrev
if match.groupdict()['hfrev'] is not None:
hfrev = int(match.groupdict()['hfrev'])
try:
branches = self._cascade[(major, minor)]
except KeyError:
LOG.debug("Ignore tag: %s", tag)
return
hf_branch = branches[HotfixBranch]
stb_branch = branches[StabilizationBranch]
dev_branch = branches[DevelopmentBranch]
if hf_branch:
if hf_branch.micro == micro:
hf_branch.hfrev = max(hfrev + 1, hf_branch.hfrev)
hf_branch.version = '%d.%d.%d.%d' % (hf_branch.major,
hf_branch.minor,
hf_branch.micro,
hf_branch.hfrev)
if stb_branch is not None and \
stb_branch.micro == hf_branch.micro:
# We have a hotfix branch but we did not remove the
# stabilization branch.
raise errors.DeprecatedStabilizationBranch(stb_branch.name,
hf_branch.name)
if stb_branch is not None and stb_branch.micro <= micro:
# We have a tag but we did not remove the stabilization branch.
raise errors.DeprecatedStabilizationBranch(stb_branch.name, tag)
if dev_branch:
dev_branch.micro = max(micro, dev_branch.micro)
|
def update_micro(self, tag):
"""Update development branch latest micro based on tag."""
pattern = r"^(?P<major>\d+)\.(?P<minor>\d+)\.(?P<micro>\d+)" \
r"(\.(?P<hfrev>\d+)|)$"
match = re.match(pattern, tag)
if not match:
LOG.debug("Ignore tag: %s", tag)
return
LOG.debug("Consider tag: %s", tag)
major = int(match.groupdict()['major'])
minor = int(match.groupdict()['minor'])
micro = int(match.groupdict()['micro'])
hfrev = 0 # default hfrev
if match.groupdict()['hfrev'] is not None:
hfrev = int(match.groupdict()['hfrev'])
try:
branches = self._cascade[(major, minor)]
except KeyError:
LOG.debug("Ignore tag: %s", tag)
return
hf_branch = branches[HotfixBranch]
stb_branch = branches[StabilizationBranch]
dev_branch = branches[DevelopmentBranch]
if hf_branch:
if hf_branch.micro == micro:
hf_branch.hfrev = max(hfrev + 1, hf_branch.hfrev)
hf_branch.version = '%d.%d.%d.%d' % (hf_branch.major,
hf_branch.minor,
hf_branch.micro,
hf_branch.hfrev)
if stb_branch is not None and \
stb_branch.micro == hf_branch.micro:
# We have a hotfix branch but we did not remove the
# stabilization branch.
raise errors.DeprecatedStabilizationBranch(stb_branch.name,
hf_branch.name)
if stb_branch is not None and stb_branch.micro <= micro:
# We have a tag but we did not remove the stabilization branch.
raise errors.DeprecatedStabilizationBranch(stb_branch.name, tag)
if dev_branch:
dev_branch.micro = max(micro, dev_branch.micro)
|
Python
|
def _set_target_versions(self, dst_branch):
"""Compute list of expected Jira FixVersion/s.
Must be called after the cascade has been finalised.
"""
for (major, minor), branch_set in self._cascade.items():
dev_branch = branch_set[DevelopmentBranch]
stb_branch = branch_set[StabilizationBranch]
hf_branch = branch_set[HotfixBranch]
if hf_branch and dst_branch.name.startswith('hotfix/'):
self.target_versions.append('%d.%d.%d.%d' % (
hf_branch.major, hf_branch.minor, hf_branch.micro,
hf_branch.hfrev))
if stb_branch:
self.target_versions.append('%d.%d.%d' % (
major, minor, stb_branch.micro))
elif dev_branch:
offset = 2 if dev_branch.has_stabilization else 1
self.target_versions.append('%d.%d.%d' % (
major, minor, dev_branch.micro + offset))
|
def _set_target_versions(self, dst_branch):
"""Compute list of expected Jira FixVersion/s.
Must be called after the cascade has been finalised.
"""
for (major, minor), branch_set in self._cascade.items():
dev_branch = branch_set[DevelopmentBranch]
stb_branch = branch_set[StabilizationBranch]
hf_branch = branch_set[HotfixBranch]
if hf_branch and dst_branch.name.startswith('hotfix/'):
self.target_versions.append('%d.%d.%d.%d' % (
hf_branch.major, hf_branch.minor, hf_branch.micro,
hf_branch.hfrev))
if stb_branch:
self.target_versions.append('%d.%d.%d' % (
major, minor, stb_branch.micro))
elif dev_branch:
offset = 2 if dev_branch.has_stabilization else 1
self.target_versions.append('%d.%d.%d' % (
major, minor, dev_branch.micro + offset))
|
Python
|
def finalize(self, dst_branch):
"""Finalize cascade considering given destination.
Assumes the cascade has been populated by calls to add_branch
and update_micro. The local lists keeping track
Args:
dst_branch: where the pull request wants to merge
Raises:
Returns:
list: list of destination branches
list: list of ignored destination branches
"""
self.get_merge_paths() # populate merge paths before removing data
ignore_stb_branches = False
include_dev_branches = False
dev_branch = None
dst_hf = dst_branch.name.startswith('hotfix/')
for (major, minor), branch_set in list(self._cascade.items()):
dev_branch = branch_set[DevelopmentBranch]
stb_branch = branch_set[StabilizationBranch]
hf_branch = branch_set[HotfixBranch]
# we have to target at least a hf or a dev branch
if dev_branch is None and hf_branch is None:
raise errors.DevBranchDoesNotExist(
'development/%d.%d' % (major, minor))
# remember if a stab is attached before it is removed
# from path, for the correct target_version computation
if stb_branch:
dev_branch.has_stabilization = True
# remove untargetted branches from cascade
if dst_branch == dev_branch:
include_dev_branches = True
ignore_stb_branches = True
if stb_branch and (ignore_stb_branches or dst_hf):
branch_set[StabilizationBranch] = None
self.ignored_branches.append(stb_branch.name)
if dst_branch == stb_branch:
include_dev_branches = True
ignore_stb_branches = True
if not include_dev_branches or dst_hf:
if branch_set[DevelopmentBranch]:
branch_set[DevelopmentBranch] = None
self.ignored_branches.append(dev_branch.name)
if branch_set[StabilizationBranch]:
branch_set[StabilizationBranch] = None
self.ignored_branches.append(stb_branch.name)
if not dst_hf:
del self._cascade[(major, minor)]
continue
if not hf_branch or hf_branch.name != dst_branch.name:
if branch_set[HotfixBranch]:
branch_set[HotfixBranch] = None
self.ignored_branches.append(hf_branch.name)
del self._cascade[(major, minor)]
# add to dst_branches in the correct order
if not dst_hf:
if branch_set[StabilizationBranch]:
self.dst_branches.append(stb_branch)
if branch_set[DevelopmentBranch]:
self.dst_branches.append(dev_branch)
else:
if branch_set[HotfixBranch]:
if dst_branch.name == hf_branch.name:
self.dst_branches.append(hf_branch)
if not dev_branch and not dst_hf:
raise errors.NotASingleDevBranch()
self._set_target_versions(dst_branch)
self.ignored_branches.sort()
|
def finalize(self, dst_branch):
"""Finalize cascade considering given destination.
Assumes the cascade has been populated by calls to add_branch
and update_micro. The local lists keeping track
Args:
dst_branch: where the pull request wants to merge
Raises:
Returns:
list: list of destination branches
list: list of ignored destination branches
"""
self.get_merge_paths() # populate merge paths before removing data
ignore_stb_branches = False
include_dev_branches = False
dev_branch = None
dst_hf = dst_branch.name.startswith('hotfix/')
for (major, minor), branch_set in list(self._cascade.items()):
dev_branch = branch_set[DevelopmentBranch]
stb_branch = branch_set[StabilizationBranch]
hf_branch = branch_set[HotfixBranch]
# we have to target at least a hf or a dev branch
if dev_branch is None and hf_branch is None:
raise errors.DevBranchDoesNotExist(
'development/%d.%d' % (major, minor))
# remember if a stab is attached before it is removed
# from path, for the correct target_version computation
if stb_branch:
dev_branch.has_stabilization = True
# remove untargetted branches from cascade
if dst_branch == dev_branch:
include_dev_branches = True
ignore_stb_branches = True
if stb_branch and (ignore_stb_branches or dst_hf):
branch_set[StabilizationBranch] = None
self.ignored_branches.append(stb_branch.name)
if dst_branch == stb_branch:
include_dev_branches = True
ignore_stb_branches = True
if not include_dev_branches or dst_hf:
if branch_set[DevelopmentBranch]:
branch_set[DevelopmentBranch] = None
self.ignored_branches.append(dev_branch.name)
if branch_set[StabilizationBranch]:
branch_set[StabilizationBranch] = None
self.ignored_branches.append(stb_branch.name)
if not dst_hf:
del self._cascade[(major, minor)]
continue
if not hf_branch or hf_branch.name != dst_branch.name:
if branch_set[HotfixBranch]:
branch_set[HotfixBranch] = None
self.ignored_branches.append(hf_branch.name)
del self._cascade[(major, minor)]
# add to dst_branches in the correct order
if not dst_hf:
if branch_set[StabilizationBranch]:
self.dst_branches.append(stb_branch)
if branch_set[DevelopmentBranch]:
self.dst_branches.append(dev_branch)
else:
if branch_set[HotfixBranch]:
if dst_branch.name == hf_branch.name:
self.dst_branches.append(hf_branch)
if not dev_branch and not dst_hf:
raise errors.NotASingleDevBranch()
self._set_target_versions(dst_branch)
self.ignored_branches.sort()
|
Python
|
def branch_factory(repo: git.Repository, branch_name: str) -> GWFBranch:
"""Construct a GWFBranch object corresponding to the branch_name.
Args:
repo: corresponding git repository.
branch_name: name of the branch to construct.
Returns:
The constructed GWFBranch.
Raises:
UnrecognizedBranchPattern if the branch name is invalid.
"""
for cls in [StabilizationBranch, DevelopmentBranch, ReleaseBranch,
QueueBranch, QueueIntegrationBranch,
FeatureBranch, HotfixBranch, LegacyHotfixBranch,
IntegrationBranch, UserBranch]:
try:
branch = cls(repo, branch_name)
return branch
except errors.BranchNameInvalid:
pass
raise errors.UnrecognizedBranchPattern(branch_name)
|
def branch_factory(repo: git.Repository, branch_name: str) -> GWFBranch:
"""Construct a GWFBranch object corresponding to the branch_name.
Args:
repo: corresponding git repository.
branch_name: name of the branch to construct.
Returns:
The constructed GWFBranch.
Raises:
UnrecognizedBranchPattern if the branch name is invalid.
"""
for cls in [StabilizationBranch, DevelopmentBranch, ReleaseBranch,
QueueBranch, QueueIntegrationBranch,
FeatureBranch, HotfixBranch, LegacyHotfixBranch,
IntegrationBranch, UserBranch]:
try:
branch = cls(repo, branch_name)
return branch
except errors.BranchNameInvalid:
pass
raise errors.UnrecognizedBranchPattern(branch_name)
|
Python
|
def build_branch_cascade(job):
"""Initialize the job's branch cascade."""
cascade = job.git.cascade
if cascade.dst_branches:
# Do not rebuild cascade
return cascade
cascade.build(job.git.repo, job.git.dst_branch)
LOG.debug(cascade.dst_branches)
return cascade
|
def build_branch_cascade(job):
"""Initialize the job's branch cascade."""
cascade = job.git.cascade
if cascade.dst_branches:
# Do not rebuild cascade
return cascade
cascade.build(job.git.repo, job.git.dst_branch)
LOG.debug(cascade.dst_branches)
return cascade
|
Python
|
def ignore_files(dir: str, files: list[str]):
"""
Returns a list of files to ignore.
To be used by shutil.copytree()
"""
return [f for f in files if Path(dir, f).is_file()]
|
def ignore_files(dir: str, files: list[str]):
"""
Returns a list of files to ignore.
To be used by shutil.copytree()
"""
return [f for f in files if Path(dir, f).is_file()]
|
Python
|
def resize_image(
image: Image.Image, max_width: int = 0, max_height: int = 0
) -> Image.Image:
"""
Resize an image and return it.
:param image: A Pill.Image object.
:param max_width: The max width the processed image can have.
:param max_height: The max height the processed image can have.
"""
old_width, old_height = image.size[0], image.size[1]
@cache
def get_proper_sizes(
max_width: int, max_height: int, old_width: int, old_height: int
) -> tuple[int, int]:
new_width, new_height = max_width, max_height
if max_width == 0:
"""If width is not set."""
new_width = round(old_width * (max_height / old_height))
elif max_height == 0:
"""If height is not set."""
new_height = round(old_height * (max_width / old_width))
else:
if old_width > old_height:
"""If image's original width is bigger than original height."""
new_height = round(old_height * (new_width / old_width))
elif old_height > old_width:
"""If image's original height is bigger than original width."""
new_width = round(old_width * (new_height / old_height))
elif old_width == old_height:
"""If image's original width and height are the same."""
if max_width > max_height:
"""If new width is bigger than new height."""
new_width = max_height
elif max_height > max_width:
"""If new height is bigger than new width."""
new_height = max_height
if new_width > max_width and max_width != 0:
new_width = max_width
new_height = round(old_height * (new_width / old_width))
if new_height > max_height and max_height != 0:
new_height = max_height
new_width = round(old_width * (new_height / old_height))
new_width = new_width
new_height = new_height
return new_width, new_height
new_width, new_height = get_proper_sizes(
max_width, max_height, old_width, old_height
)
image = image.resize((new_width, new_height), Image.LANCZOS)
return image
|
def resize_image(
image: Image.Image, max_width: int = 0, max_height: int = 0
) -> Image.Image:
"""
Resize an image and return it.
:param image: A Pill.Image object.
:param max_width: The max width the processed image can have.
:param max_height: The max height the processed image can have.
"""
old_width, old_height = image.size[0], image.size[1]
@cache
def get_proper_sizes(
max_width: int, max_height: int, old_width: int, old_height: int
) -> tuple[int, int]:
new_width, new_height = max_width, max_height
if max_width == 0:
"""If width is not set."""
new_width = round(old_width * (max_height / old_height))
elif max_height == 0:
"""If height is not set."""
new_height = round(old_height * (max_width / old_width))
else:
if old_width > old_height:
"""If image's original width is bigger than original height."""
new_height = round(old_height * (new_width / old_width))
elif old_height > old_width:
"""If image's original height is bigger than original width."""
new_width = round(old_width * (new_height / old_height))
elif old_width == old_height:
"""If image's original width and height are the same."""
if max_width > max_height:
"""If new width is bigger than new height."""
new_width = max_height
elif max_height > max_width:
"""If new height is bigger than new width."""
new_height = max_height
if new_width > max_width and max_width != 0:
new_width = max_width
new_height = round(old_height * (new_width / old_width))
if new_height > max_height and max_height != 0:
new_height = max_height
new_width = round(old_width * (new_height / old_height))
new_width = new_width
new_height = new_height
return new_width, new_height
new_width, new_height = get_proper_sizes(
max_width, max_height, old_width, old_height
)
image = image.resize((new_width, new_height), Image.LANCZOS)
return image
|
Python
|
def dir_generator() -> tuple[Path, Path, Path]:
"""Generates files and folders for the test."""
parent_dir = Path(tmp_path, "parent")
first_child = Path(parent_dir, "first_child")
second_child = Path(first_child, "second_child")
parent_dir.mkdir()
first_child.mkdir()
second_child.mkdir()
tmp_img = Image.new("RGB", (original_width, original_height))
for index, dir in enumerate((parent_dir, first_child, second_child)):
for i in range(index + 1):
img_path = Path(dir, f"image-{i}.jpg")
tmp_img.save(img_path)
return parent_dir, first_child, second_child
|
def dir_generator() -> tuple[Path, Path, Path]:
"""Generates files and folders for the test."""
parent_dir = Path(tmp_path, "parent")
first_child = Path(parent_dir, "first_child")
second_child = Path(first_child, "second_child")
parent_dir.mkdir()
first_child.mkdir()
second_child.mkdir()
tmp_img = Image.new("RGB", (original_width, original_height))
for index, dir in enumerate((parent_dir, first_child, second_child)):
for i in range(index + 1):
img_path = Path(dir, f"image-{i}.jpg")
tmp_img.save(img_path)
return parent_dir, first_child, second_child
|
Python
|
def SanityCheckName(self,name):
"""The way to check that the name is good.
For now we only allow letters, numbers and underscores.
The name is used directly as part of a file-name and should therefore be nice."""
import re
if not re.match("""[a-zA-Z0-9_]+""", name):
raise self.NameError(name)
|
def SanityCheckName(self,name):
"""The way to check that the name is good.
For now we only allow letters, numbers and underscores.
The name is used directly as part of a file-name and should therefore be nice."""
import re
if not re.match("""[a-zA-Z0-9_]+""", name):
raise self.NameError(name)
|
Python
|
def SessionName(self,name, nocheck=False):
"""The only way anyone whould ever contruct the session name"""
if not nocheck:
self.SanityCheckName(name)
return self.session_dir + os.sep + self.session_pattern % name
|
def SessionName(self,name, nocheck=False):
"""The only way anyone whould ever contruct the session name"""
if not nocheck:
self.SanityCheckName(name)
return self.session_dir + os.sep + self.session_pattern % name
|
Python
|
def StateName(self,name, nocheck=False):
"""The only way anyone should ever construct the state name"""
if not nocheck:
self.SanityCheckName(name)
return self.session_dir + os.sep + self.state_pattern % name
|
def StateName(self,name, nocheck=False):
"""The only way anyone should ever construct the state name"""
if not nocheck:
self.SanityCheckName(name)
return self.session_dir + os.sep + self.state_pattern % name
|
Python
|
def ConstructDifference(self, old_env, new_env ):
"""Construct the dictionary of EWDiffObjects that is called a 'state'."""
from EnvironmentObjects import EWDiffObject
diff = {}
#consider all keys that are in either einvironment, old or new
for key in set(new_env.keys() + old_env.keys()):
if key not in new_env:
#it must be in old_env
diff[key] = EWDiffObject(old=old_env[key])
# new is none, this corresponds to a removal
elif key not in old_env:
# it must be in new_env
diff[key] = EWDiffObject(new=new_env[key])
# old is none, this is an addition
else:
# maybe nothing changed? This is most likely.
if old_env[key] != new_env[key]:
# something changed
diff[key] = EWDiffObject(old=old_env[key], new=new_env[key])
return diff
|
def ConstructDifference(self, old_env, new_env ):
"""Construct the dictionary of EWDiffObjects that is called a 'state'."""
from EnvironmentObjects import EWDiffObject
diff = {}
#consider all keys that are in either einvironment, old or new
for key in set(new_env.keys() + old_env.keys()):
if key not in new_env:
#it must be in old_env
diff[key] = EWDiffObject(old=old_env[key])
# new is none, this corresponds to a removal
elif key not in old_env:
# it must be in new_env
diff[key] = EWDiffObject(new=new_env[key])
# old is none, this is an addition
else:
# maybe nothing changed? This is most likely.
if old_env[key] != new_env[key]:
# something changed
diff[key] = EWDiffObject(old=old_env[key], new=new_env[key])
return diff
|
Python
|
def GetStateOrSessionDiff(self,name):
import os
log("Session Dir",os.listdir(self.session_dir))
"""If an existing saved state by this name is found, return that,if an open session by this name is found, construct the difference to the present."""
session_file_name = self.SessionName(name)
state_file_name = self.StateName(name)
try:
diff = self.GetState(name)
except:
log()
log("session wans't available. making it.")
try:
diff = self.ConstructDifference(self.GetSession(name), self.shell.environment)
except:
log()
log("that didn't work either")
raise self.UnknownName(name)
return diff
|
def GetStateOrSessionDiff(self,name):
import os
log("Session Dir",os.listdir(self.session_dir))
"""If an existing saved state by this name is found, return that,if an open session by this name is found, construct the difference to the present."""
session_file_name = self.SessionName(name)
state_file_name = self.StateName(name)
try:
diff = self.GetState(name)
except:
log()
log("session wans't available. making it.")
try:
diff = self.ConstructDifference(self.GetSession(name), self.shell.environment)
except:
log()
log("that didn't work either")
raise self.UnknownName(name)
return diff
|
Python
|
def CloseSession(self,name):
"""Close an existing open recording session"""
oldsession = self.GetSession(name)
diff = self.ConstructDifference(oldsession, self.shell.environment)
self.WriteState(name, diff)
# Remove the session file, thereby closing the session
import os
os.remove(self.SessionName(name))
|
def CloseSession(self,name):
"""Close an existing open recording session"""
oldsession = self.GetSession(name)
diff = self.ConstructDifference(oldsession, self.shell.environment)
self.WriteState(name, diff)
# Remove the session file, thereby closing the session
import os
os.remove(self.SessionName(name))
|
Python
|
def undo(self,name):
"""remove the changes associated with a state."""
import os
if not os.path.exists(self.StateName(name)) and os.path.exists(self.SessionName(name)):
force = ("-f" in sys.argv or "--force" in sys.argv)
if force:
self.ClearState(name)
self.CloseSession(name)
else:
print >>sys.stderr, "Session %s is open. Do you want to close it? " % name,
user=raw_input()
log("Question:","Session %s is open. Do you want to close it? " % name,"answer:",user)
if re.match("""y(es?)?""",user):
print >>sys.stderr," --> Closing."
self.UserClearState(name)
self.CloseSession(name)
else:
print >>sys.stderr," --> Using open session."
return self.redo(name,Reverse=True)
|
def undo(self,name):
"""remove the changes associated with a state."""
import os
if not os.path.exists(self.StateName(name)) and os.path.exists(self.SessionName(name)):
force = ("-f" in sys.argv or "--force" in sys.argv)
if force:
self.ClearState(name)
self.CloseSession(name)
else:
print >>sys.stderr, "Session %s is open. Do you want to close it? " % name,
user=raw_input()
log("Question:","Session %s is open. Do you want to close it? " % name,"answer:",user)
if re.match("""y(es?)?""",user):
print >>sys.stderr," --> Closing."
self.UserClearState(name)
self.CloseSession(name)
else:
print >>sys.stderr," --> Using open session."
return self.redo(name,Reverse=True)
|
Python
|
def GetShellParser(shell_var):
"""Decides which shell environment reader and diff writer to use."""
if shell_var == "/bin/bash":
from Bash import BashInteractor
return BashInteractor
else:
raise Exception("Unknown shell.")
|
def GetShellParser(shell_var):
"""Decides which shell environment reader and diff writer to use."""
if shell_var == "/bin/bash":
from Bash import BashInteractor
return BashInteractor
else:
raise Exception("Unknown shell.")
|
Python
|
def _crop_center(image):
"""Returns a cropped square image."""
shape = image.shape
new_shape = min(shape[1], shape[2])
offset_y = max(shape[1] - shape[2], 0) // 2
offset_x = max(shape[2] - shape[1], 0) // 2
return tf.image.crop_to_bounding_box(image, offset_y, offset_x, new_shape, new_shape)
|
def _crop_center(image):
"""Returns a cropped square image."""
shape = image.shape
new_shape = min(shape[1], shape[2])
offset_y = max(shape[1] - shape[2], 0) // 2
offset_x = max(shape[2] - shape[1], 0) // 2
return tf.image.crop_to_bounding_box(image, offset_y, offset_x, new_shape, new_shape)
|
Python
|
def fseq(options):
'''
Call peaks using F-seq
'''
# parse options
if not which('fseq'):
sys.exit('Error: No F-seq installed!')
folder = check_dir(options['<rampagedir>'])
flength = options['-l']
wig_flag = options['--wig']
percent = float(options['-p'])
with open(os.path.join(folder, 'total_counts.txt'), 'r') as f:
total = int(f.read().rstrip())
# run F-seq
flist = {'+': 'rampage_plus_5end.bed', '-': 'rampage_minus_5end.bed'}
all_peak_f = os.path.join(folder, 'rampage_peaks.txt')
with open(all_peak_f, 'w') as out:
for strand in flist:
peak_f = run_fseq(folder, flist[strand], strand, flength, wig_flag,
percent)
with open(peak_f, 'r') as f:
for line in f:
if total: # calculate RPM
reads = int(line.rstrip().split()[9])
rpm = reads * 1000000.0 / total
out.write(line.rstrip() + '\t%f\n' % rpm)
else:
out.write(line)
|
def fseq(options):
'''
Call peaks using F-seq
'''
# parse options
if not which('fseq'):
sys.exit('Error: No F-seq installed!')
folder = check_dir(options['<rampagedir>'])
flength = options['-l']
wig_flag = options['--wig']
percent = float(options['-p'])
with open(os.path.join(folder, 'total_counts.txt'), 'r') as f:
total = int(f.read().rstrip())
# run F-seq
flist = {'+': 'rampage_plus_5end.bed', '-': 'rampage_minus_5end.bed'}
all_peak_f = os.path.join(folder, 'rampage_peaks.txt')
with open(all_peak_f, 'w') as out:
for strand in flist:
peak_f = run_fseq(folder, flist[strand], strand, flength, wig_flag,
percent)
with open(peak_f, 'r') as f:
for line in f:
if total: # calculate RPM
reads = int(line.rstrip().split()[9])
rpm = reads * 1000000.0 / total
out.write(line.rstrip() + '\t%f\n' % rpm)
else:
out.write(line)
|
Python
|
def forward(self, support, support_mask, query, query_mask):
"""Att-Induction Networks forward.
Args:
support: torch.Tensor, [-1, N, K, max_length]
support_mask: torch.Tensor, [-1, N, K, max_length]
query: torch.Tensor, [-1, totalQ, max_length]
query_mask: torch.Tensor, [-1, totalQ, max_length]
Returns:
relation_score: torch.Tensor, [B, totalQ, N]
predict_label: torch.Tensor, [B, totalQ]"""
B, N, K = support.size()[:3]
totalQ = query.size()[1] # Number of query instances for each batch
# 1. Encoder
support = support.view(-1, self.max_length) # [B * N * K, max_length]
support_mask = support_mask.view(-1, self.max_length)
query = query.view(-1, self.max_length) # [B * totalQ, max_length]
query_mask = query_mask.view(-1, self.max_length)
support = self.encoder(support, support_mask) # [B * N * K, D]
query = self.encoder(query, query_mask) # [B * totalQ, D]
support = support.view(-1, N, K, self.hidden_size) # [B, N, K, D]
query = query.view(-1, totalQ, self.hidden_size) # [B, totalQ, D]
# 2. Induction
# 2.1 Attention score
support_att = support.unsqueeze(1).expand(-1, totalQ, -1, -1, -1) # [B, totalQ, N, K, D]
query_att = query.unsqueeze(2).unsqueeze(3).expand(-1, -1, N, -1, -1) # [B, totalQ, N, 1, D]
support_query_att = torch.cat((query_att, support_att), dim=3) # [B, totalQ, N, 1 + K, D]
support_query_att = support_query_att.view(-1, 1 + K, self.hidden_size).transpose(0, 1) # [1 + K, B * totalQ * N , D]
att_score = self.self_att(support_query_att, support_query_att, support_query_att)[0] # [1 + K, B * totalQ * N , D]
att_score = support_query_att + self.dropout(att_score)
att_score = self.layer_norm(att_score)
att_score = att_score[0].view(-1, totalQ, N, self.hidden_size) # [B, totalQ, N, D]
att_score = att_score.unsqueeze(3) # [B, totalQ, N, 1, D]
# 2.2 Attention-based dynamic routing
support_hat = self.__squash(self.fc_induction(support).unsqueeze(1).expand(-1, totalQ, -1, -1, -1)) # [B, totalQ, N, K, D]
b = torch.zeros(B, totalQ, N, K, 1, device=self.current_device, requires_grad=False) # [B, totalQ, N, K, 1]
for _ in range(self.induction_iters):
d = F.softmax(b, dim=3) # [B, totalQ, N, K, 1]
c_hat = torch.mul(d, support_hat).sum(3, keepdims=True) # [B, totalQ, N, 1, D]
c = self.__squash(c_hat) # [B, totalQ, N, 1, D]
b = b + torch.mul(att_score, torch.tanh(torch.mul(support_hat, c))).sum(-1, keepdims=True) # [B, totalQ, N, K, 1]
# 3. Relation
c = c.squeeze(3) # [B, totalQ, N, D]
query = query.unsqueeze(2).expand(-1, -1, N, -1) # [B, totalQ, N, D]
query = query.contiguous()
relation_score = self.relation(c, query) # [B, totalQ, N]
predict_label = relation_score.argmax(dim=-1, keepdims=False) # [B, totalQ]
return relation_score, predict_label
|
def forward(self, support, support_mask, query, query_mask):
"""Att-Induction Networks forward.
Args:
support: torch.Tensor, [-1, N, K, max_length]
support_mask: torch.Tensor, [-1, N, K, max_length]
query: torch.Tensor, [-1, totalQ, max_length]
query_mask: torch.Tensor, [-1, totalQ, max_length]
Returns:
relation_score: torch.Tensor, [B, totalQ, N]
predict_label: torch.Tensor, [B, totalQ]"""
B, N, K = support.size()[:3]
totalQ = query.size()[1] # Number of query instances for each batch
# 1. Encoder
support = support.view(-1, self.max_length) # [B * N * K, max_length]
support_mask = support_mask.view(-1, self.max_length)
query = query.view(-1, self.max_length) # [B * totalQ, max_length]
query_mask = query_mask.view(-1, self.max_length)
support = self.encoder(support, support_mask) # [B * N * K, D]
query = self.encoder(query, query_mask) # [B * totalQ, D]
support = support.view(-1, N, K, self.hidden_size) # [B, N, K, D]
query = query.view(-1, totalQ, self.hidden_size) # [B, totalQ, D]
# 2. Induction
# 2.1 Attention score
support_att = support.unsqueeze(1).expand(-1, totalQ, -1, -1, -1) # [B, totalQ, N, K, D]
query_att = query.unsqueeze(2).unsqueeze(3).expand(-1, -1, N, -1, -1) # [B, totalQ, N, 1, D]
support_query_att = torch.cat((query_att, support_att), dim=3) # [B, totalQ, N, 1 + K, D]
support_query_att = support_query_att.view(-1, 1 + K, self.hidden_size).transpose(0, 1) # [1 + K, B * totalQ * N , D]
att_score = self.self_att(support_query_att, support_query_att, support_query_att)[0] # [1 + K, B * totalQ * N , D]
att_score = support_query_att + self.dropout(att_score)
att_score = self.layer_norm(att_score)
att_score = att_score[0].view(-1, totalQ, N, self.hidden_size) # [B, totalQ, N, D]
att_score = att_score.unsqueeze(3) # [B, totalQ, N, 1, D]
# 2.2 Attention-based dynamic routing
support_hat = self.__squash(self.fc_induction(support).unsqueeze(1).expand(-1, totalQ, -1, -1, -1)) # [B, totalQ, N, K, D]
b = torch.zeros(B, totalQ, N, K, 1, device=self.current_device, requires_grad=False) # [B, totalQ, N, K, 1]
for _ in range(self.induction_iters):
d = F.softmax(b, dim=3) # [B, totalQ, N, K, 1]
c_hat = torch.mul(d, support_hat).sum(3, keepdims=True) # [B, totalQ, N, 1, D]
c = self.__squash(c_hat) # [B, totalQ, N, 1, D]
b = b + torch.mul(att_score, torch.tanh(torch.mul(support_hat, c))).sum(-1, keepdims=True) # [B, totalQ, N, K, 1]
# 3. Relation
c = c.squeeze(3) # [B, totalQ, N, D]
query = query.unsqueeze(2).expand(-1, -1, N, -1) # [B, totalQ, N, D]
query = query.contiguous()
relation_score = self.relation(c, query) # [B, totalQ, N]
predict_label = relation_score.argmax(dim=-1, keepdims=False) # [B, totalQ]
return relation_score, predict_label
|
Python
|
def forward(self, tokens, lengths):
"""Self-Attention Bi-LSTM encoder forward.
Args:
tokens: torch.Tensor, [-1, max_length]
lengths: torch.Tensor, [-1, max_length]
Returns:
sentence_embedding: torch.Tensor, [-1, 2 * hidden_size]"""
embedding = self.word_embedding(tokens) # [-1, max_length, word_vec_dim]
lengths = lengths[:, 0] # [-1]
embedding = nn.utils.rnn.pack_padded_sequence(
embedding,
lengths,
batch_first=True,
enforce_sorted=False
)
output, _ = self.bilstm(embedding) # packed sequence. Throw away (h_n, c_n)
output, _ = nn.utils.rnn.pad_packed_sequence(
output,
batch_first=True,
padding_value=0.0,
total_length=self.max_length
) # [-1, max_length, 2 * hidden_size]. Throw away lengths tensor
att_score = torch.tanh(self.att1(output)) # [-1, max_length, att_dim]
att_score = F.softmax(self.att2(att_score), dim=1) # [-1, max_length, 1]
output = torch.mul(att_score, output).sum(dim=1, keepdim=False) # [-1, 2 * hidden_size]
return output
|
def forward(self, tokens, lengths):
"""Self-Attention Bi-LSTM encoder forward.
Args:
tokens: torch.Tensor, [-1, max_length]
lengths: torch.Tensor, [-1, max_length]
Returns:
sentence_embedding: torch.Tensor, [-1, 2 * hidden_size]"""
embedding = self.word_embedding(tokens) # [-1, max_length, word_vec_dim]
lengths = lengths[:, 0] # [-1]
embedding = nn.utils.rnn.pack_padded_sequence(
embedding,
lengths,
batch_first=True,
enforce_sorted=False
)
output, _ = self.bilstm(embedding) # packed sequence. Throw away (h_n, c_n)
output, _ = nn.utils.rnn.pad_packed_sequence(
output,
batch_first=True,
padding_value=0.0,
total_length=self.max_length
) # [-1, max_length, 2 * hidden_size]. Throw away lengths tensor
att_score = torch.tanh(self.att1(output)) # [-1, max_length, att_dim]
att_score = F.softmax(self.att2(att_score), dim=1) # [-1, max_length, 1]
output = torch.mul(att_score, output).sum(dim=1, keepdim=False) # [-1, 2 * hidden_size]
return output
|
Python
|
def glove_preprocessing(word_vec_file, output_path):
"""Transforming English word embedding txt into .npy embedding matrix and JSON index file."""
token2idx = {}
word_vec = []
with open(word_vec_file, "r") as f:
line = f.readline()
index = 0
while line:
line = line.strip().split()
token, vec = line[0], line[1:]
vec = list(map(float, vec))
if token in token2idx:
raise ValueError("{} is existed!".format(token))
else:
token2idx[token] = index
word_vec.append(vec)
index += 1
line = f.readline()
word_vec = np.array(word_vec)
assert len(token2idx) == np.shape(word_vec)[0], "Length is not same!"
json.dump(token2idx, open(os.path.join(output_path, "token2idx.json"), "w"))
np.save(os.path.join(output_path, "word_vec.npy"), word_vec)
|
def glove_preprocessing(word_vec_file, output_path):
"""Transforming English word embedding txt into .npy embedding matrix and JSON index file."""
token2idx = {}
word_vec = []
with open(word_vec_file, "r") as f:
line = f.readline()
index = 0
while line:
line = line.strip().split()
token, vec = line[0], line[1:]
vec = list(map(float, vec))
if token in token2idx:
raise ValueError("{} is existed!".format(token))
else:
token2idx[token] = index
word_vec.append(vec)
index += 1
line = f.readline()
word_vec = np.array(word_vec)
assert len(token2idx) == np.shape(word_vec)[0], "Length is not same!"
json.dump(token2idx, open(os.path.join(output_path, "token2idx.json"), "w"))
np.save(os.path.join(output_path, "word_vec.npy"), word_vec)
|
Python
|
def chinese_word_vec_preprocessing(word_vec_file, output_path):
"""Transforming Chinese word embedding txt into .npy embedding matrix and JSON index file."""
token2idx = {}
word_vec = []
with open(word_vec_file, "r") as f:
line = f.readline() # 1292607 300
line = f.readline()
index = 0
while line:
line = line.rstrip().split(" ")
# print(line)
token, vec = line[0], line[1:]
vec = list(map(float, vec))
if token in token2idx:
print("{} is existed!".format(token))
line = f.readline()
continue
else:
token2idx[token] = index
word_vec.append(vec)
index += 1
line = f.readline()
if index % 100000 == 0:
print("{:d} done!".format(index))
word_vec = np.array(word_vec)
assert len(token2idx) == np.shape(word_vec)[0], "Length is not same!"
json.dump(token2idx, open(os.path.join(output_path, "token2idx.json"), "w"))
np.save(os.path.join(output_path, "word_vec.npy"), word_vec)
|
def chinese_word_vec_preprocessing(word_vec_file, output_path):
"""Transforming Chinese word embedding txt into .npy embedding matrix and JSON index file."""
token2idx = {}
word_vec = []
with open(word_vec_file, "r") as f:
line = f.readline() # 1292607 300
line = f.readline()
index = 0
while line:
line = line.rstrip().split(" ")
# print(line)
token, vec = line[0], line[1:]
vec = list(map(float, vec))
if token in token2idx:
print("{} is existed!".format(token))
line = f.readline()
continue
else:
token2idx[token] = index
word_vec.append(vec)
index += 1
line = f.readline()
if index % 100000 == 0:
print("{:d} done!".format(index))
word_vec = np.array(word_vec)
assert len(token2idx) == np.shape(word_vec)[0], "Length is not same!"
json.dump(token2idx, open(os.path.join(output_path, "token2idx.json"), "w"))
np.save(os.path.join(output_path, "word_vec.npy"), word_vec)
|
Python
|
def make_train_indices(test_indices, n_examples, downsample=False, labels=None):
"""Makes a set of train indices from a set of test indices.
test_indices: List of test indices.
n_examples: Number of total (testing + training) examples.
downsample: Whether to downsample. Default False.
labels: Binary labels. Only needed if downsample is True.
-> sorted list of training indices.
"""
if downsample and labels is None:
raise ValueError('labels must be specified if downsample is specified')
training = sorted(set(range(n_examples)) - set(test_indices))
if not downsample:
logging.debug('Not downsampling...')
return training
logging.debug('Downsampling...')
training = numpy.array(training)
positives = training[labels[training] == 1]
negatives = training[labels[training] == 0]
if positives.shape[0] > negatives.shape[0]:
numpy.random.shuffle(positives)
positives = positives[:negatives.shape[0]]
training = numpy.concatenate([positives, negatives])
else:
numpy.random.shuffle(negatives)
negatives = negatives[:positives.shape[0]]
training = numpy.concatenate([negatives, positives])
training.sort()
logging.debug('Positive examples: {}'.format(positives.shape[0]))
logging.debug('Negative examples: {}'.format(negatives.shape[0]))
logging.debug('Total examples: {}'.format(training.shape[0]))
assert training.shape[0] == positives.shape[0] + negatives.shape[0]
return training
|
def make_train_indices(test_indices, n_examples, downsample=False, labels=None):
"""Makes a set of train indices from a set of test indices.
test_indices: List of test indices.
n_examples: Number of total (testing + training) examples.
downsample: Whether to downsample. Default False.
labels: Binary labels. Only needed if downsample is True.
-> sorted list of training indices.
"""
if downsample and labels is None:
raise ValueError('labels must be specified if downsample is specified')
training = sorted(set(range(n_examples)) - set(test_indices))
if not downsample:
logging.debug('Not downsampling...')
return training
logging.debug('Downsampling...')
training = numpy.array(training)
positives = training[labels[training] == 1]
negatives = training[labels[training] == 0]
if positives.shape[0] > negatives.shape[0]:
numpy.random.shuffle(positives)
positives = positives[:negatives.shape[0]]
training = numpy.concatenate([positives, negatives])
else:
numpy.random.shuffle(negatives)
negatives = negatives[:positives.shape[0]]
training = numpy.concatenate([negatives, positives])
training.sort()
logging.debug('Positive examples: {}'.format(positives.shape[0]))
logging.debug('Negative examples: {}'.format(negatives.shape[0]))
logging.debug('Total examples: {}'.format(training.shape[0]))
assert training.shape[0] == positives.shape[0] + negatives.shape[0]
return training
|
Python
|
def lr(results, method_name, split_id, features, targets, test_indices, C=1.0,
overwrite=False, n_train=None, selection='passive'):
"""Run logistic regression and store results.
results: Results object.
features: (n_examples, n_features) array of features.
targets: (n_examples,) array of binary targets.
test_indices: List of integer testing indices.
method_name: Name of this method in the results.
split_id: ID of the split in the results.
C: Regularisation parameter. Default 1.0. Higher values mean less
regularisation.
overwrite: Whether to overwrite existing results (default False).
n_train: Number of training examples. Default as many as possible.
selection: How to select training examples. Default 'passive'.
"""
assert max(test_indices) < features.shape[0]
assert min(test_indices) >= 0
if results.has_run(method_name, split_id) and not overwrite:
logging.info('Skipping trial {}:{}.'.format(method_name, split_id))
return
train_indices = make_train_indices(test_indices, features.shape[0])
if n_train is not None and n_train != len(train_indices):
assert n_train <= len(train_indices)
train_indices, _ = sklearn.cross_validation.train_test_split(
train_indices, train_size=n_train, stratify=targets[train_indices])
train_indices.sort()
lr = sklearn.linear_model.LogisticRegression(class_weight='balanced', C=C,
n_jobs=-1)
logging.debug('Fitting {} instances.'.format(len(features)))
lr.fit(features[train_indices], targets[train_indices])
logging.debug('Fit complete. Storing...'.format(len(features)))
results.store_trial(
method_name, split_id,
lr.predict_proba(features[test_indices])[:, 1],
indices=test_indices, params=lr_to_params(lr))
|
def lr(results, method_name, split_id, features, targets, test_indices, C=1.0,
overwrite=False, n_train=None, selection='passive'):
"""Run logistic regression and store results.
results: Results object.
features: (n_examples, n_features) array of features.
targets: (n_examples,) array of binary targets.
test_indices: List of integer testing indices.
method_name: Name of this method in the results.
split_id: ID of the split in the results.
C: Regularisation parameter. Default 1.0. Higher values mean less
regularisation.
overwrite: Whether to overwrite existing results (default False).
n_train: Number of training examples. Default as many as possible.
selection: How to select training examples. Default 'passive'.
"""
assert max(test_indices) < features.shape[0]
assert min(test_indices) >= 0
if results.has_run(method_name, split_id) and not overwrite:
logging.info('Skipping trial {}:{}.'.format(method_name, split_id))
return
train_indices = make_train_indices(test_indices, features.shape[0])
if n_train is not None and n_train != len(train_indices):
assert n_train <= len(train_indices)
train_indices, _ = sklearn.cross_validation.train_test_split(
train_indices, train_size=n_train, stratify=targets[train_indices])
train_indices.sort()
lr = sklearn.linear_model.LogisticRegression(class_weight='balanced', C=C,
n_jobs=-1)
logging.debug('Fitting {} instances.'.format(len(features)))
lr.fit(features[train_indices], targets[train_indices])
logging.debug('Fit complete. Storing...'.format(len(features)))
results.store_trial(
method_name, split_id,
lr.predict_proba(features[test_indices])[:, 1],
indices=test_indices, params=lr_to_params(lr))
|
Python
|
def rf(results, method_name, split_id, features, targets, test_indices,
overwrite=False):
"""Run random forest and store results.
Does not store the model as random forests are difficult to serialise.
results: Results object.
features: (n_examples, n_features) array of features.
targets: (n_examples,) array of binary targets.
test_indices: List of integer testing indices.
method_name: Name of this method in the results.
split_id: ID of the split in the results.
overwrite: Whether to overwrite existing results (default False).
"""
assert max(test_indices) < features.shape[0]
assert min(test_indices) >= 0
train_indices = make_train_indices(test_indices, features.shape[0])
if results.has_run(method_name, split_id) and not overwrite:
logging.info('Skipping trial {}:{}.'.format(method_name, split_id))
return
rf = sklearn.ensemble.RandomForestClassifier(class_weight='balanced')
rf.fit(features[train_indices], targets[train_indices])
results.store_trial(
method_name, split_id,
rf.predict_proba(features[test_indices])[:, 1],
indices=test_indices, params=numpy.zeros((results.n_params,)))
|
def rf(results, method_name, split_id, features, targets, test_indices,
overwrite=False):
"""Run random forest and store results.
Does not store the model as random forests are difficult to serialise.
results: Results object.
features: (n_examples, n_features) array of features.
targets: (n_examples,) array of binary targets.
test_indices: List of integer testing indices.
method_name: Name of this method in the results.
split_id: ID of the split in the results.
overwrite: Whether to overwrite existing results (default False).
"""
assert max(test_indices) < features.shape[0]
assert min(test_indices) >= 0
train_indices = make_train_indices(test_indices, features.shape[0])
if results.has_run(method_name, split_id) and not overwrite:
logging.info('Skipping trial {}:{}.'.format(method_name, split_id))
return
rf = sklearn.ensemble.RandomForestClassifier(class_weight='balanced')
rf.fit(features[train_indices], targets[train_indices])
results.store_trial(
method_name, split_id,
rf.predict_proba(features[test_indices])[:, 1],
indices=test_indices, params=numpy.zeros((results.n_params,)))
|
Python
|
def raykar(results, method_name, split_id, features, targets, test_indices,
overwrite=False, n_restarts=5, downsample=False):
"""Run the Raykar algorithm and store results.
results: Results object.
features: (n_examples, n_features) array of features.
targets: (n_labellers, n_examples) masked array of binary targets.
test_indices: List of integer testing indices.
method_name: Name of this method in the results.
split_id: ID of the split in the results.
overwrite: Whether to overwrite existing results (default False).
n_restarts: Number of random restarts. Default 5.
downsample: Whether to downsample. Default False.
"""
assert max(test_indices) < features.shape[0]
assert min(test_indices) >= 0
train_indices = make_train_indices(test_indices, features.shape[0],
downsample=downsample,
labels=majority_vote(targets))
if results.has_run(method_name, split_id) and not overwrite:
logging.info('Skipping trial {}:{}.'.format(method_name, split_id))
return
rc = RaykarClassifier(n_restarts=n_restarts)
rc.fit(features[train_indices], targets[:, train_indices])
results.store_trial(
method_name, split_id,
rc.predict_proba(features[test_indices]),
indices=test_indices, params=rc.serialise())
|
def raykar(results, method_name, split_id, features, targets, test_indices,
overwrite=False, n_restarts=5, downsample=False):
"""Run the Raykar algorithm and store results.
results: Results object.
features: (n_examples, n_features) array of features.
targets: (n_labellers, n_examples) masked array of binary targets.
test_indices: List of integer testing indices.
method_name: Name of this method in the results.
split_id: ID of the split in the results.
overwrite: Whether to overwrite existing results (default False).
n_restarts: Number of random restarts. Default 5.
downsample: Whether to downsample. Default False.
"""
assert max(test_indices) < features.shape[0]
assert min(test_indices) >= 0
train_indices = make_train_indices(test_indices, features.shape[0],
downsample=downsample,
labels=majority_vote(targets))
if results.has_run(method_name, split_id) and not overwrite:
logging.info('Skipping trial {}:{}.'.format(method_name, split_id))
return
rc = RaykarClassifier(n_restarts=n_restarts)
rc.fit(features[train_indices], targets[:, train_indices])
results.store_trial(
method_name, split_id,
rc.predict_proba(features[test_indices]),
indices=test_indices, params=rc.serialise())
|
Python
|
def yan(results, method_name, split_id, features, targets, test_indices,
overwrite=False, n_restarts=5):
"""Run the Yan algorithm and store results.
results: Results object.
features: (n_examples, n_features) array of features.
targets: (n_labellers, n_examples) masked array of binary targets.
test_indices: List of integer testing indices.
method_name: Name of this method in the results.
split_id: ID of the split in the results.
overwrite: Whether to overwrite existing results (default False).
n_restarts: Number of random restarts. Default 5.
"""
assert max(test_indices) < features.shape[0]
assert min(test_indices) >= 0
train_indices = make_train_indices(test_indices, features.shape[0])
if results.has_run(method_name, split_id) and not overwrite:
logging.info('Skipping trial {}:{}.'.format(method_name, split_id))
return
yc = YanClassifier(n_restarts=n_restarts)
yc.fit(features[train_indices], targets[:, train_indices])
results.store_trial(method_name, split_id,
yc.predict_proba(features[test_indices]),
indices=test_indices, params=yc.serialise())
|
def yan(results, method_name, split_id, features, targets, test_indices,
overwrite=False, n_restarts=5):
"""Run the Yan algorithm and store results.
results: Results object.
features: (n_examples, n_features) array of features.
targets: (n_labellers, n_examples) masked array of binary targets.
test_indices: List of integer testing indices.
method_name: Name of this method in the results.
split_id: ID of the split in the results.
overwrite: Whether to overwrite existing results (default False).
n_restarts: Number of random restarts. Default 5.
"""
assert max(test_indices) < features.shape[0]
assert min(test_indices) >= 0
train_indices = make_train_indices(test_indices, features.shape[0])
if results.has_run(method_name, split_id) and not overwrite:
logging.info('Skipping trial {}:{}.'.format(method_name, split_id))
return
yc = YanClassifier(n_restarts=n_restarts)
yc.fit(features[train_indices], targets[:, train_indices])
results.store_trial(method_name, split_id,
yc.predict_proba(features[test_indices]),
indices=test_indices, params=yc.serialise())
|
Python
|
def generate(f_h5, out_f_h5, field='cdfs'):
"""Generates potential hosts and their astronomical features.
f_h5: crowdastro input HDF5 file.
out_f_h5: Training data output HDF5 file.
"""
ir_survey = f_h5.attrs['ir_survey']
if ir_survey == 'swire':
swire = f_h5['/swire/{}/numeric'.format(field)]
fluxes = swire[:, 2:7]
# Skip stellarities.
distances = swire[:, 8].reshape((-1, 1))
images = swire[:, 9:]
coords = swire[:, :2]
s1_s2 = fluxes[:, 0] / fluxes[:, 1]
s2_s3 = fluxes[:, 1] / fluxes[:, 2]
s3_s4 = fluxes[:, 2] / fluxes[:, 3]
astro_features = numpy.concatenate(
[fluxes,
s1_s2.reshape((-1, 1)),
s2_s3.reshape((-1, 1)),
s3_s4.reshape((-1, 1))], axis=1)
elif ir_survey == 'wise':
wise = f_h5['/wise/{}/numeric'.format(field)]
magnitudes = wise[:, 2:6]
distances = wise[:, 6].reshape((-1, 1))
images = wise[:, 7:]
coords = wise[:, :2]
# Magnitude differences are probably useful features.
w1_w2 = magnitudes[:, 0] - magnitudes[:, 1]
w2_w3 = magnitudes[:, 1] - magnitudes[:, 2]
# Converting the magnitudes to a linear scale seems to improve
# performance.
linearised_magnitudes = numpy.power(10, -0.4 * magnitudes)
w1_w2 = numpy.power(10, -0.4 * w1_w2)
w2_w3 = numpy.power(10, -0.4 * w2_w3)
astro_features = numpy.concatenate(
[linearised_magnitudes,
w1_w2.reshape((-1, 1)),
w2_w3.reshape((-1, 1))], axis=1)
n_features = config['surveys'][ir_survey]['n_features']
assert astro_features.shape[1] + distances.shape[1] == n_features
# We now need to find the labels for each.
if field == 'cdfs':
truths = set(f_h5['/atlas/cdfs/consensus_objects'][:, 1])
labels = numpy.array([o in truths for o in range(len(astro_features))])
assert len(labels) == len(astro_features)
out_f_h5.create_dataset('labels', data=labels)
assert len(astro_features) == len(distances)
assert len(distances) == len(images)
features = numpy.hstack([astro_features, distances, images])
n_astro = features.shape[1] - images.shape[1]
# Save to HDF5.
out_f_h5.create_dataset('raw_features', data=features)
out_f_h5.create_dataset('positions', data=coords)
out_f_h5.attrs['ir_survey'] = ir_survey
out_f_h5.attrs['field'] = field
|
def generate(f_h5, out_f_h5, field='cdfs'):
"""Generates potential hosts and their astronomical features.
f_h5: crowdastro input HDF5 file.
out_f_h5: Training data output HDF5 file.
"""
ir_survey = f_h5.attrs['ir_survey']
if ir_survey == 'swire':
swire = f_h5['/swire/{}/numeric'.format(field)]
fluxes = swire[:, 2:7]
# Skip stellarities.
distances = swire[:, 8].reshape((-1, 1))
images = swire[:, 9:]
coords = swire[:, :2]
s1_s2 = fluxes[:, 0] / fluxes[:, 1]
s2_s3 = fluxes[:, 1] / fluxes[:, 2]
s3_s4 = fluxes[:, 2] / fluxes[:, 3]
astro_features = numpy.concatenate(
[fluxes,
s1_s2.reshape((-1, 1)),
s2_s3.reshape((-1, 1)),
s3_s4.reshape((-1, 1))], axis=1)
elif ir_survey == 'wise':
wise = f_h5['/wise/{}/numeric'.format(field)]
magnitudes = wise[:, 2:6]
distances = wise[:, 6].reshape((-1, 1))
images = wise[:, 7:]
coords = wise[:, :2]
# Magnitude differences are probably useful features.
w1_w2 = magnitudes[:, 0] - magnitudes[:, 1]
w2_w3 = magnitudes[:, 1] - magnitudes[:, 2]
# Converting the magnitudes to a linear scale seems to improve
# performance.
linearised_magnitudes = numpy.power(10, -0.4 * magnitudes)
w1_w2 = numpy.power(10, -0.4 * w1_w2)
w2_w3 = numpy.power(10, -0.4 * w2_w3)
astro_features = numpy.concatenate(
[linearised_magnitudes,
w1_w2.reshape((-1, 1)),
w2_w3.reshape((-1, 1))], axis=1)
n_features = config['surveys'][ir_survey]['n_features']
assert astro_features.shape[1] + distances.shape[1] == n_features
# We now need to find the labels for each.
if field == 'cdfs':
truths = set(f_h5['/atlas/cdfs/consensus_objects'][:, 1])
labels = numpy.array([o in truths for o in range(len(astro_features))])
assert len(labels) == len(astro_features)
out_f_h5.create_dataset('labels', data=labels)
assert len(astro_features) == len(distances)
assert len(distances) == len(images)
features = numpy.hstack([astro_features, distances, images])
n_astro = features.shape[1] - images.shape[1]
# Save to HDF5.
out_f_h5.create_dataset('raw_features', data=features)
out_f_h5.create_dataset('positions', data=coords)
out_f_h5.attrs['ir_survey'] = ir_survey
out_f_h5.attrs['field'] = field
|
Python
|
def compute_disagreement(self):
"""Finds disagreement for all objects in the pool."""
labels = numpy.array([c.predict(self.pool) for c in self.classifiers])
# Each column is the classifications of one data point.
n_agree = labels.sum(axis=0)
self.disagreement = numpy.ma.masked_array(
numpy.abs(n_agree - self.n_classifiers // 2),
mask=~self.labels.mask)
|
def compute_disagreement(self):
"""Finds disagreement for all objects in the pool."""
labels = numpy.array([c.predict(self.pool) for c in self.classifiers])
# Each column is the classifications of one data point.
n_agree = labels.sum(axis=0)
self.disagreement = numpy.ma.masked_array(
numpy.abs(n_agree - self.n_classifiers // 2),
mask=~self.labels.mask)
|
Python
|
def score(self, test_xs, test_ts):
"""Finds cross-entropy error on test data."""
return numpy.mean([
sklearn.metrics.log_loss(
test_ts, c.predict(test_xs))
for c in self.classifiers])
|
def score(self, test_xs, test_ts):
"""Finds cross-entropy error on test data."""
return numpy.mean([
sklearn.metrics.log_loss(
test_ts, c.predict(test_xs))
for c in self.classifiers])
|
Python
|
def ba(self, test_xs, test_ts):
"""Finds balanced accuracy on test data."""
# labels = numpy.zeros((self.n_classifiers, len(test_xs)))
# for c in range(self.n_classifiers):
# preds = self.classifiers[c].predict(test_xs)
# labels[c, :] = preds
# labels = numpy.ma.MaskedArray(labels, mask=numpy.zeros(labels.shape))
# return balanced_accuracy(test_ts, majority_vote(labels))
return balanced_accuracy(
test_ts, self.reference_classifier.predict(test_xs))
|
def ba(self, test_xs, test_ts):
"""Finds balanced accuracy on test data."""
# labels = numpy.zeros((self.n_classifiers, len(test_xs)))
# for c in range(self.n_classifiers):
# preds = self.classifiers[c].predict(test_xs)
# labels[c, :] = preds
# labels = numpy.ma.MaskedArray(labels, mask=numpy.zeros(labels.shape))
# return balanced_accuracy(test_ts, majority_vote(labels))
return balanced_accuracy(
test_ts, self.reference_classifier.predict(test_xs))
|
Python
|
def unpack(params, n_dim, n_annotators):
"""Unpacks an array of parameters into a, b, and h."""
a = params[:n_dim]
b = params[n_dim]
h = params[n_dim + 1:]
assert h.shape[0] == n_annotators
return a, b, h
|
def unpack(params, n_dim, n_annotators):
"""Unpacks an array of parameters into a, b, and h."""
a = params[:n_dim]
b = params[n_dim]
h = params[n_dim + 1:]
assert h.shape[0] == n_annotators
return a, b, h
|
Python
|
def sample_index(self):
"""Finds index of a random unlabelled point."""
unlabelled = self.labels.mask.nonzero()[0]
if len(unlabelled):
index = numpy.random.choice(unlabelled)
return index
return 0
|
def sample_index(self):
"""Finds index of a random unlabelled point."""
unlabelled = self.labels.mask.nonzero()[0]
if len(unlabelled):
index = numpy.random.choice(unlabelled)
return index
return 0
|
Python
|
def sample_indices(self, n):
"""Finds indices of n random unlabelled points."""
indices = set()
unlabelled = self.labels.mask.nonzero()[0]
if len(unlabelled) < n:
return unlabelled
while len(indices) < n:
index = numpy.random.choice(unlabelled)
indices.add(index)
return sorted(indices)
|
def sample_indices(self, n):
"""Finds indices of n random unlabelled points."""
indices = set()
unlabelled = self.labels.mask.nonzero()[0]
if len(unlabelled) < n:
return unlabelled
while len(indices) < n:
index = numpy.random.choice(unlabelled)
indices.add(index)
return sorted(indices)
|
Python
|
def sample_index(self):
"""Finds index of a random unlabelled point."""
unlabelled = self.labels.mask.nonzero()[0]
unlabelled_groundtruth = self.labels.data[unlabelled]
if len(unlabelled):
if numpy.random.random() < 0.5:
index = numpy.random.choice(
unlabelled[unlabelled_groundtruth == 1])
else:
index = numpy.random.choice(
unlabelled[unlabelled_groundtruth == 0])
return index
return 0
|
def sample_index(self):
"""Finds index of a random unlabelled point."""
unlabelled = self.labels.mask.nonzero()[0]
unlabelled_groundtruth = self.labels.data[unlabelled]
if len(unlabelled):
if numpy.random.random() < 0.5:
index = numpy.random.choice(
unlabelled[unlabelled_groundtruth == 1])
else:
index = numpy.random.choice(
unlabelled[unlabelled_groundtruth == 0])
return index
return 0
|
Python
|
def sample_indices(self, n):
"""Finds indices of n random unlabelled points."""
indices = set()
unlabelled = self.labels.mask.nonzero()[0]
if len(unlabelled) < n:
return unlabelled
unlabelled_groundtruth = self.labels.data[unlabelled]
while len(indices) < n:
if ((numpy.random.random() < 0.5 and
len(unlabelled[unlabelled_groundtruth == 1]) > 0) or
len(unlabelled[unlabelled_groundtruth == 0]) == 0):
index = numpy.random.choice(
unlabelled[unlabelled_groundtruth == 1])
else:
index = numpy.random.choice(
unlabelled[unlabelled_groundtruth == 0])
indices.add(index)
return sorted(indices)
|
def sample_indices(self, n):
"""Finds indices of n random unlabelled points."""
indices = set()
unlabelled = self.labels.mask.nonzero()[0]
if len(unlabelled) < n:
return unlabelled
unlabelled_groundtruth = self.labels.data[unlabelled]
while len(indices) < n:
if ((numpy.random.random() < 0.5 and
len(unlabelled[unlabelled_groundtruth == 1]) > 0) or
len(unlabelled[unlabelled_groundtruth == 0]) == 0):
index = numpy.random.choice(
unlabelled[unlabelled_groundtruth == 1])
else:
index = numpy.random.choice(
unlabelled[unlabelled_groundtruth == 0])
indices.add(index)
return sorted(indices)
|
Python
|
def compute_uncertainties(self):
"""Finds uncertainties for all objects in the pool."""
# To keep things simple, I'll use the (negative) proximity to the
# decision boundary as the uncertainty. Note that the uncertainties
# array is masked such that labelled points have no uncertainty.
probs = self.classifier.predict_proba(self.pool)[:, 1]
self.uncertainties = numpy.ma.masked_array(-numpy.abs(probs - 0.5),
mask=~self.labels.mask)
|
def compute_uncertainties(self):
"""Finds uncertainties for all objects in the pool."""
# To keep things simple, I'll use the (negative) proximity to the
# decision boundary as the uncertainty. Note that the uncertainties
# array is masked such that labelled points have no uncertainty.
probs = self.classifier.predict_proba(self.pool)[:, 1]
self.uncertainties = numpy.ma.masked_array(-numpy.abs(probs - 0.5),
mask=~self.labels.mask)
|
Python
|
def _top_n_accurate_targets(crowdastro_h5, y_true, n_annotators=5,
threshold=700):
"""Get the labels of the top n most accurate annotators assessed against
y_true, above a threshold number of annotations."""
ir_survey = crowdastro_h5.attrs['ir_survey']
labels = crowdastro_h5['/{}/cdfs/rgz_raw_labels'.format(ir_survey)].value
labels_mask = crowdastro_h5[
'/{}/cdfs/rgz_raw_labels_mask'.format(ir_survey)].value
labels = numpy.ma.MaskedArray(labels, mask=labels_mask)
# Compare each annotator to the Norris labels. Get their balanced accuracy.
annotator_accuracies = []
for t in range(labels.shape[0]):
cm = sklearn.metrics.confusion_matrix(y_true[~labels[t].mask],
labels[t][~labels[t].mask])
if cm.shape[0] == 1:
continue
tp = cm[1, 1]
n, p = cm.sum(axis=1)
tn = cm[0, 0]
if not n or not p or p + n < threshold:
annotator_accuracies.append(0)
continue
ba = (tp / p + tn / n) / 2
annotator_accuracies.append(ba)
ranked_annotators = numpy.argsort(annotator_accuracies)
top_n_annotators = ranked_annotators[-n_annotators:]
return labels[top_n_annotators]
|
def _top_n_accurate_targets(crowdastro_h5, y_true, n_annotators=5,
threshold=700):
"""Get the labels of the top n most accurate annotators assessed against
y_true, above a threshold number of annotations."""
ir_survey = crowdastro_h5.attrs['ir_survey']
labels = crowdastro_h5['/{}/cdfs/rgz_raw_labels'.format(ir_survey)].value
labels_mask = crowdastro_h5[
'/{}/cdfs/rgz_raw_labels_mask'.format(ir_survey)].value
labels = numpy.ma.MaskedArray(labels, mask=labels_mask)
# Compare each annotator to the Norris labels. Get their balanced accuracy.
annotator_accuracies = []
for t in range(labels.shape[0]):
cm = sklearn.metrics.confusion_matrix(y_true[~labels[t].mask],
labels[t][~labels[t].mask])
if cm.shape[0] == 1:
continue
tp = cm[1, 1]
n, p = cm.sum(axis=1)
tn = cm[0, 0]
if not n or not p or p + n < threshold:
annotator_accuracies.append(0)
continue
ba = (tp / p + tn / n) / 2
annotator_accuracies.append(ba)
ranked_annotators = numpy.argsort(annotator_accuracies)
top_n_annotators = ranked_annotators[-n_annotators:]
return labels[top_n_annotators]
|
Python
|
def top_n_accurate_targets(crowdastro_h5, n_annotators=5, threshold=700):
"""Get the labels of the top n most accurate annotators, assessed
against the groundtruth, above a threshold number of annotations.
"""
ir_survey = crowdastro_h5.attrs['ir_survey']
norris = crowdastro_h5['/{}/cdfs/norris_labels'.format(ir_survey)].value
return _top_n_accurate_targets(crowdastro_h5, norris,
n_annotators=n_annotators,
threshold=threshold)
|
def top_n_accurate_targets(crowdastro_h5, n_annotators=5, threshold=700):
"""Get the labels of the top n most accurate annotators, assessed
against the groundtruth, above a threshold number of annotations.
"""
ir_survey = crowdastro_h5.attrs['ir_survey']
norris = crowdastro_h5['/{}/cdfs/norris_labels'.format(ir_survey)].value
return _top_n_accurate_targets(crowdastro_h5, norris,
n_annotators=n_annotators,
threshold=threshold)
|
Python
|
def top_n_mv_accurate_targets(crowdastro_h5, n_annotators=5, threshold=700):
"""Get the labels of the top n most accurate annotators, assessed
against the majority vote, above a threshold number of annotations.
"""
ir_survey = crowdastro_h5.attrs['ir_survey']
labels = crowdastro_h5['/{}/cdfs/rgz_raw_labels'.format(ir_survey)].value
labels_mask = crowdastro_h5[
'/{}/cdfs/rgz_raw_labels_mask'.format(ir_survey)].value
labels = numpy.ma.MaskedArray(labels, mask=labels_mask)
mv = majority_vote(labels)
return _top_n_accurate_targets(crowdastro_h5, mv,
n_annotators=n_annotators,
threshold=threshold)
|
def top_n_mv_accurate_targets(crowdastro_h5, n_annotators=5, threshold=700):
"""Get the labels of the top n most accurate annotators, assessed
against the majority vote, above a threshold number of annotations.
"""
ir_survey = crowdastro_h5.attrs['ir_survey']
labels = crowdastro_h5['/{}/cdfs/rgz_raw_labels'.format(ir_survey)].value
labels_mask = crowdastro_h5[
'/{}/cdfs/rgz_raw_labels_mask'.format(ir_survey)].value
labels = numpy.ma.MaskedArray(labels, mask=labels_mask)
mv = majority_vote(labels)
return _top_n_accurate_targets(crowdastro_h5, mv,
n_annotators=n_annotators,
threshold=threshold)
|
Python
|
def top_n_prolific_targets(crowdastro_h5, n_annotators=5):
"""Get the labels of the top n most prolific annotators."""
labels = crowdastro_h5['/wise/cdfs/rgz_raw_labels'].value
labels_mask = crowdastro_h5['/wise/cdfs/rgz_raw_labels_mask'].value
labels = numpy.ma.MaskedArray(labels, mask=labels_mask)
n_seen = labels_mask.sum(axis=1)
top_seen = numpy.argsort(n_seen)[-n_annotators:]
return labels[top_seen]
|
def top_n_prolific_targets(crowdastro_h5, n_annotators=5):
"""Get the labels of the top n most prolific annotators."""
labels = crowdastro_h5['/wise/cdfs/rgz_raw_labels'].value
labels_mask = crowdastro_h5['/wise/cdfs/rgz_raw_labels_mask'].value
labels = numpy.ma.MaskedArray(labels, mask=labels_mask)
n_seen = labels_mask.sum(axis=1)
top_seen = numpy.argsort(n_seen)[-n_annotators:]
return labels[top_seen]
|
Python
|
def _unpack(self, params, n_dim, n_annotators):
"""Unpacks an array of parameters into a and w."""
a = params[:n_dim]
w = params[n_dim:].reshape((n_annotators, n_dim))
return a, w
|
def _unpack(self, params, n_dim, n_annotators):
"""Unpacks an array of parameters into a and w."""
a = params[:n_dim]
w = params[n_dim:].reshape((n_annotators, n_dim))
return a, w
|
Python
|
def predict(self, x):
"""Classify data points using logistic regression.
x: Data points. (n_samples, n_dim) NumPy array.
"""
return numpy.round(self.predict_proba(x))
|
def predict(self, x):
"""Classify data points using logistic regression.
x: Data points. (n_samples, n_dim) NumPy array.
"""
return numpy.round(self.predict_proba(x))
|
Python
|
def predict_proba(self, x):
"""Predict probabilities of data points using logistic regression.
x: Data points. (n_samples, n_dim) NumPy array.
"""
x = numpy.hstack([x, numpy.ones((x.shape[0], 1))])
return logistic_regression(self.a_, x)
|
def predict_proba(self, x):
"""Predict probabilities of data points using logistic regression.
x: Data points. (n_samples, n_dim) NumPy array.
"""
x = numpy.hstack([x, numpy.ones((x.shape[0], 1))])
return logistic_regression(self.a_, x)
|
Python
|
def score(self, X, Y):
"""Computes the likelihood of labels and data under the model.
X: (n_samples, n_features) NumPy array of data.
Y: (n_labellers, n_samples) NumPy masked array of crowd labels.
"""
X = numpy.hstack([X, numpy.ones((X.shape[0], 1))])
return self._likelihood(self.a_, self.w_, X, Y)
|
def score(self, X, Y):
"""Computes the likelihood of labels and data under the model.
X: (n_samples, n_features) NumPy array of data.
Y: (n_labellers, n_samples) NumPy masked array of crowd labels.
"""
X = numpy.hstack([X, numpy.ones((X.shape[0], 1))])
return self._likelihood(self.a_, self.w_, X, Y)
|
Python
|
def _likelihood(self, a, w, X, Y):
"""Computes the likelihood of labels and data under a model.
X: (n_samples, n_features) NumPy array of data.
Y: (n_labellers, n_samples) NumPy masked array of crowd labels.
"""
lh = 1
for i in range(X.shape[0]):
for t in range(Y.shape[0]):
if Y.mask[t, i]:
continue
lr = logistic_regression(a, X[i])
p1 = self._annotator_model(w[t], X[i], Y[t, i], 1) * lr
p0 = self._annotator_model(w[t], X[i], Y[t, i], 0) * (1 - lr)
lh *= p1 + p0
return lh
|
def _likelihood(self, a, w, X, Y):
"""Computes the likelihood of labels and data under a model.
X: (n_samples, n_features) NumPy array of data.
Y: (n_labellers, n_samples) NumPy masked array of crowd labels.
"""
lh = 1
for i in range(X.shape[0]):
for t in range(Y.shape[0]):
if Y.mask[t, i]:
continue
lr = logistic_regression(a, X[i])
p1 = self._annotator_model(w[t], X[i], Y[t, i], 1) * lr
p0 = self._annotator_model(w[t], X[i], Y[t, i], 0) * (1 - lr)
lh *= p1 + p0
return lh
|
Python
|
def require_atlas(f):
"""Decorator that ensures a subject (the first argument) is from the ATLAS
survey.
"""
def g(subject, *args, **kwargs):
if subject['metadata']['survey'] != 'atlas':
raise ValueError('Subject not from ATLAS survey.')
return f(subject, *args, **kwargs)
return g
|
def require_atlas(f):
"""Decorator that ensures a subject (the first argument) is from the ATLAS
survey.
"""
def g(subject, *args, **kwargs):
if subject['metadata']['survey'] != 'atlas':
raise ValueError('Subject not from ATLAS survey.')
return f(subject, *args, **kwargs)
return g
|
Python
|
def open_fits(subject, field, wavelength, size='2x2'):
"""Opens a FITS image of a subject.
Can be used as a context handler.
subject: RGZ subject dict, from the ATLAS survey.
field: 'elais' or 'cdfs'
wavelength: 'ir' or 'radio'
size: Optional. '2x2' or '5x5'.
-> FITS image file.
"""
if field not in {'elais-s1', 'cdfs'}:
raise ValueError('field must be either "elais-s1" or "cdfs".')
if wavelength not in {'ir', 'radio'}:
raise ValueError('wavelength must be either "ir" or "radio".')
cid = subject['metadata']['source']
filename = '{}_{}.fits'.format(cid, wavelength)
path = os.path.join(config['data_sources']['{}_fits'.format(field)], size,
filename)
return astropy.io.fits.open(path, ignore_blank=True)
|
def open_fits(subject, field, wavelength, size='2x2'):
"""Opens a FITS image of a subject.
Can be used as a context handler.
subject: RGZ subject dict, from the ATLAS survey.
field: 'elais' or 'cdfs'
wavelength: 'ir' or 'radio'
size: Optional. '2x2' or '5x5'.
-> FITS image file.
"""
if field not in {'elais-s1', 'cdfs'}:
raise ValueError('field must be either "elais-s1" or "cdfs".')
if wavelength not in {'ir', 'radio'}:
raise ValueError('wavelength must be either "ir" or "radio".')
cid = subject['metadata']['source']
filename = '{}_{}.fits'.format(cid, wavelength)
path = os.path.join(config['data_sources']['{}_fits'.format(field)], size,
filename)
return astropy.io.fits.open(path, ignore_blank=True)
|
Python
|
def train(x, y, epsilon=1e-5, lr_init=False, skip_zeros=False):
"""Expectation-maximisation algorithm from Yan et al. (2010).
x: Data. (n_samples, n_dim) NumPy array
y: Labels. (n_annotators, n_samples) NumPy array
epsilon: Convergence threshold. Default 1e-5. float
lr_init: Initialised with logistic regression. Default False.
skip_zeros: Whether to detect and skip zero probabilities. Default False.
"""
# TODO(MatthewJA): Restore skip_zeros functionality.
n_samples, n_dim = x.shape
n_annotators, n_samples_ = y.shape
assert n_samples == n_samples_, 'Label array has wrong number of labels.'
# Compute majority vote labels (for debugging + logistic regression init).
majority_y = numpy.zeros((n_samples,))
for i in range(n_samples):
labels = y[:, i]
if labels.mask is False:
counter = collections.Counter(labels)
else:
counter = collections.Counter(labels[~labels.mask])
if counter:
majority_y[i] = max(counter, key=counter.get)
else:
# No labels for this data point.
majority_y[i] = numpy.random.randint(2) # ¯\_(ツ)_/¯
logging.info('Initialising...')
if lr_init:
# For our initial guess, we'll fit logistic regression to the majority
# vote.
lr_ab = sklearn.linear_model.LogisticRegression()
lr_ab.fit(x, majority_y)
a = lr_ab.coef_.ravel()
b = lr_ab.intercept_[0]
else:
a = numpy.random.normal(size=(n_dim,))
b = numpy.random.normal()
w = numpy.random.normal(size=(n_annotators, n_dim))
g = numpy.ones((n_annotators,))
logging.debug('Initial a: %s', a)
logging.debug('Initial b: %s', b)
logging.debug('Initial w: %s', w)
logging.debug('Initial g: %s', g)
assert x.shape == (n_samples, n_dim)
assert y.shape == (n_annotators, n_samples)
logging.info('Iterating until convergence...')
iters = 0
while True: # Until convergence (checked later).
iters += 1
logging.info('Iteration %d', iters)
a_, b_, w_, g_ = em_step(
n_samples, n_annotators, n_dim, a, b, w, g, x, y)
# Check convergence.
dist = numpy.linalg.norm(a - a_) ** 2 + (b - b_) ** 2
logging.debug('Distance: {:.02f}'.format(dist))
if dist <= epsilon:
return a_, b_, w_, g_
a, b, w, g = a_, b_, w_, g_
|
def train(x, y, epsilon=1e-5, lr_init=False, skip_zeros=False):
"""Expectation-maximisation algorithm from Yan et al. (2010).
x: Data. (n_samples, n_dim) NumPy array
y: Labels. (n_annotators, n_samples) NumPy array
epsilon: Convergence threshold. Default 1e-5. float
lr_init: Initialised with logistic regression. Default False.
skip_zeros: Whether to detect and skip zero probabilities. Default False.
"""
# TODO(MatthewJA): Restore skip_zeros functionality.
n_samples, n_dim = x.shape
n_annotators, n_samples_ = y.shape
assert n_samples == n_samples_, 'Label array has wrong number of labels.'
# Compute majority vote labels (for debugging + logistic regression init).
majority_y = numpy.zeros((n_samples,))
for i in range(n_samples):
labels = y[:, i]
if labels.mask is False:
counter = collections.Counter(labels)
else:
counter = collections.Counter(labels[~labels.mask])
if counter:
majority_y[i] = max(counter, key=counter.get)
else:
# No labels for this data point.
majority_y[i] = numpy.random.randint(2) # ¯\_(ツ)_/¯
logging.info('Initialising...')
if lr_init:
# For our initial guess, we'll fit logistic regression to the majority
# vote.
lr_ab = sklearn.linear_model.LogisticRegression()
lr_ab.fit(x, majority_y)
a = lr_ab.coef_.ravel()
b = lr_ab.intercept_[0]
else:
a = numpy.random.normal(size=(n_dim,))
b = numpy.random.normal()
w = numpy.random.normal(size=(n_annotators, n_dim))
g = numpy.ones((n_annotators,))
logging.debug('Initial a: %s', a)
logging.debug('Initial b: %s', b)
logging.debug('Initial w: %s', w)
logging.debug('Initial g: %s', g)
assert x.shape == (n_samples, n_dim)
assert y.shape == (n_annotators, n_samples)
logging.info('Iterating until convergence...')
iters = 0
while True: # Until convergence (checked later).
iters += 1
logging.info('Iteration %d', iters)
a_, b_, w_, g_ = em_step(
n_samples, n_annotators, n_dim, a, b, w, g, x, y)
# Check convergence.
dist = numpy.linalg.norm(a - a_) ** 2 + (b - b_) ** 2
logging.debug('Distance: {:.02f}'.format(dist))
if dist <= epsilon:
return a_, b_, w_, g_
a, b, w, g = a_, b_, w_, g_
|
Python
|
def run_sample(data_path, catalogue_path, limit=0):
"""Runs the consensus algorithm described in Banfield et al., 2015."""
# TODO(MatthewJA): Reimplement the 'update' argument.
# TODO(MatthewJA): Reimplement the 'do_plot' argument.
# TODO(MatthewJA): This only works for ATLAS subjects.
# TODO(MatthewJA): The original code only worked on FIRST subjects. Have I
# broken anything by running ATLAS subjects through it?
paths = load_contours.make_pathdict(data_path, catalogue_path)
sample_subjects = [cz for cz in subjects.find({
'metadata.survey': 'atlas'
}).limit(limit)]
logging.debug('Found {} subjects.'.format(len(sample_subjects)))
zooniverse_ids = [cz['zooniverse_id'] for cz in sample_subjects]
with open('%s/csv/%s.csv' % (rgz_dir, PATH_STEM), 'w') as csv_file:
writer = csv.writer(csv_file)
writer.writerow([
'zooniverse_id',
'first_id',
'n_users',
'n_total',
'consensus_level',
'n_radio',
'label',
'bbox',
'ir_peak',
])
for idx, zid in enumerate(zooniverse_ids):
logging.debug('Zooniverse ID: {}'.format(zid))
if not idx % 100:
# Progress.
now = datetime.datetime.now().strftime('%H:%M:%S.%f')
progress = idx / len(zooniverse_ids)
logging.info('{:.02%} {}'.format(progress, now))
cons = consensus(zid)
cons['consensus_level'] = cons['n_users'] / cons['n_total']
# CSV.
for ans in cons['answer'].values():
try:
ir_peak = ans['ir_peak']
except KeyError:
ir_peak = ans.get('ir', None)
writer.writerow([
cons['zid'],
cons['source'],
cons['n_users'],
cons['n_total'],
cons['consensus_level'],
len(ans['xmax']),
alpha(ans['ind']),
bbox_unravel(ans['bbox']),
ir_peak,
])
# CSV
cmaster = pandas.read_csv('%s/csv/%s.csv' % (rgz_dir, PATH_STEM))
cmaster75 = cmaster[cmaster['consensus_level'] >= 0.75]
cmaster75.to_csv('%s/csv/%s_75.csv' % (rgz_dir, PATH_STEM), index=False)
logging.info('Completed consensus.')
|
def run_sample(data_path, catalogue_path, limit=0):
"""Runs the consensus algorithm described in Banfield et al., 2015."""
# TODO(MatthewJA): Reimplement the 'update' argument.
# TODO(MatthewJA): Reimplement the 'do_plot' argument.
# TODO(MatthewJA): This only works for ATLAS subjects.
# TODO(MatthewJA): The original code only worked on FIRST subjects. Have I
# broken anything by running ATLAS subjects through it?
paths = load_contours.make_pathdict(data_path, catalogue_path)
sample_subjects = [cz for cz in subjects.find({
'metadata.survey': 'atlas'
}).limit(limit)]
logging.debug('Found {} subjects.'.format(len(sample_subjects)))
zooniverse_ids = [cz['zooniverse_id'] for cz in sample_subjects]
with open('%s/csv/%s.csv' % (rgz_dir, PATH_STEM), 'w') as csv_file:
writer = csv.writer(csv_file)
writer.writerow([
'zooniverse_id',
'first_id',
'n_users',
'n_total',
'consensus_level',
'n_radio',
'label',
'bbox',
'ir_peak',
])
for idx, zid in enumerate(zooniverse_ids):
logging.debug('Zooniverse ID: {}'.format(zid))
if not idx % 100:
# Progress.
now = datetime.datetime.now().strftime('%H:%M:%S.%f')
progress = idx / len(zooniverse_ids)
logging.info('{:.02%} {}'.format(progress, now))
cons = consensus(zid)
cons['consensus_level'] = cons['n_users'] / cons['n_total']
# CSV.
for ans in cons['answer'].values():
try:
ir_peak = ans['ir_peak']
except KeyError:
ir_peak = ans.get('ir', None)
writer.writerow([
cons['zid'],
cons['source'],
cons['n_users'],
cons['n_total'],
cons['consensus_level'],
len(ans['xmax']),
alpha(ans['ind']),
bbox_unravel(ans['bbox']),
ir_peak,
])
# CSV
cmaster = pandas.read_csv('%s/csv/%s.csv' % (rgz_dir, PATH_STEM))
cmaster75 = cmaster[cmaster['consensus_level'] >= 0.75]
cmaster75.to_csv('%s/csv/%s_75.csv' % (rgz_dir, PATH_STEM), index=False)
logging.info('Completed consensus.')
|
Python
|
def hash_file(f):
"""Finds the MD5 hash of a file.
File must be opened in bytes mode.
"""
h = hashlib.md5()
chunk_size = 65536 # 64 KiB
for chunk in iter(lambda: f.read(chunk_size), b''):
h.update(chunk)
return h.hexdigest()
|
def hash_file(f):
"""Finds the MD5 hash of a file.
File must be opened in bytes mode.
"""
h = hashlib.md5()
chunk_size = 65536 # 64 KiB
for chunk in iter(lambda: f.read(chunk_size), b''):
h.update(chunk)
return h.hexdigest()
|
Python
|
def prep_h5(f_h5, ir_survey):
"""Creates hierarchy in HDF5 file."""
f_h5.create_group('/atlas/cdfs')
f_h5.create_group('/atlas/elais')
f_h5.create_group('/{}/cdfs'.format(ir_survey))
f_h5.create_group('/{}/elais'.format(ir_survey))
f_h5.attrs['version'] = VERSION
f_h5.attrs['ir_survey'] = ir_survey
|
def prep_h5(f_h5, ir_survey):
"""Creates hierarchy in HDF5 file."""
f_h5.create_group('/atlas/cdfs')
f_h5.create_group('/atlas/elais')
f_h5.create_group('/{}/cdfs'.format(ir_survey))
f_h5.create_group('/{}/elais'.format(ir_survey))
f_h5.attrs['version'] = VERSION
f_h5.attrs['ir_survey'] = ir_survey
|
Python
|
def import_atlas(f_h5, test=False, field='cdfs'):
"""Imports the ATLAS dataset into crowdastro, as well as associated SWIRE.
f_h5: An HDF5 file.
test: Flag to run on only 10 subjects. Default False.
"""
from . import rgz_data as data
# Fetch groups from HDF5.
cdfs = f_h5['/atlas/{}'.format(field)]
# First pass, I'll find coords, names, and Zooniverse IDs, as well as how
# many data points there are.
coords = []
names = []
zooniverse_ids = []
if (field == 'cdfs'):
# We need the ATLAS name, but we can only get it by going through the
# ATLAS catalogue and finding the nearest component.
# https://github.com/chengsoonong/crowdastro/issues/63
# Fortunately, @jbanfield has already done this, so we can just load
# that CSV and match the names.
# TODO(MatthewJA): This matches the ATLAS component ID, but maybe we
# should be using the name instead.
rgz_to_atlas = {}
with open(config['data_sources']['rgz_to_atlas']) as f:
reader = csv.DictReader(f)
for row in reader:
rgz_to_atlas[row['ID_RGZ']] = row['ID']
all_subjects = data.get_all_subjects(survey='atlas', field=field)
if test:
all_subjects = all_subjects.limit(10)
for subject in all_subjects:
ra, dec = subject['coords']
zooniverse_id = subject['zooniverse_id']
rgz_source_id = subject['metadata']['source']
if rgz_source_id not in rgz_to_atlas:
logging.debug('Skipping %s; no matching ATLAS component.',
zooniverse_id)
continue
name = rgz_to_atlas[rgz_source_id]
# Store the results.
coords.append((ra, dec))
names.append(name)
zooniverse_ids.append(zooniverse_id)
elif (field == 'elais'):
atlascatalogue = ascii.read(config['data_sources']['atlas_catalogue'])
ras, decs = atlascatalogue['RA_deg'], atlascatalogue['Dec_deg']
e_ids = atlascatalogue['ID']
fields = atlascatalogue['field']
# Store the results.
for ra, dec, e_id, field_ in zip(ras, decs, e_ids, fields):
if (field_ == 'ELAIS-S1'):
coords.append((ra, dec))
names.append(e_id)
zooniverse_ids.append(e_id)
n_cdfs = len(names)
# Sort the data by Zooniverse ID.
coords_to_zooniverse_ids = dict(zip(coords, zooniverse_ids))
names_to_zooniverse_ids = dict(zip(names, zooniverse_ids))
coords.sort(key=coords_to_zooniverse_ids.get)
names.sort(key=names_to_zooniverse_ids.get)
zooniverse_ids.sort()
# Begin to store the data. We will have two tables: one for numeric data,
# and one for strings. We will have to preallocate the numeric table so that
# we aren't storing huge amounts of image data in memory.
# Strings.
dtype = [('zooniverse_id', '<S{}'.format(MAX_ZOONIVERSE_ID_LENGTH)),
('name', '<S{}'.format(MAX_NAME_LENGTH))]
string_data = numpy.array(list(zip(zooniverse_ids, names)), dtype=dtype)
cdfs.create_dataset('string', data=string_data, dtype=dtype)
# Numeric.
image_size = (config['surveys']['atlas']['fits_width'] *
config['surveys']['atlas']['fits_height'])
# RA, DEC, radio, (distance to SWIRE object added later)
dim = (n_cdfs, 1 + 1 + image_size)
numeric = cdfs.create_dataset('_numeric', shape=dim, dtype='float32')
# Load image patches and store numeric data.
with astropy.io.fits.open(
config['data_sources']['atlas_{}_image'.format(field)],
ignore_blank=True) as atlas_image:
wcs = astropy.wcs.WCS(atlas_image[0].header).dropaxis(3).dropaxis(2)
pix_coords = wcs.all_world2pix(coords, FITS_CONVENTION)
assert pix_coords.shape[1] == 2
logging.debug('Fetching %d ATLAS images.', len(pix_coords))
for index, (x, y) in enumerate(pix_coords):
radio = atlas_image[0].data[
0, 0, # stokes, freq
int(y) - config['surveys']['atlas']['fits_height'] // 2:
int(y) + config['surveys']['atlas']['fits_height'] // 2,
int(x) - config['surveys']['atlas']['fits_width'] // 2:
int(x) + config['surveys']['atlas']['fits_width'] // 2]
numeric[index, 0] = coords[index][0]
numeric[index, 1] = coords[index][1]
numeric[index, 2:2 + image_size] = radio.reshape(-1)
logging.debug('ATLAS imported.')
|
def import_atlas(f_h5, test=False, field='cdfs'):
"""Imports the ATLAS dataset into crowdastro, as well as associated SWIRE.
f_h5: An HDF5 file.
test: Flag to run on only 10 subjects. Default False.
"""
from . import rgz_data as data
# Fetch groups from HDF5.
cdfs = f_h5['/atlas/{}'.format(field)]
# First pass, I'll find coords, names, and Zooniverse IDs, as well as how
# many data points there are.
coords = []
names = []
zooniverse_ids = []
if (field == 'cdfs'):
# We need the ATLAS name, but we can only get it by going through the
# ATLAS catalogue and finding the nearest component.
# https://github.com/chengsoonong/crowdastro/issues/63
# Fortunately, @jbanfield has already done this, so we can just load
# that CSV and match the names.
# TODO(MatthewJA): This matches the ATLAS component ID, but maybe we
# should be using the name instead.
rgz_to_atlas = {}
with open(config['data_sources']['rgz_to_atlas']) as f:
reader = csv.DictReader(f)
for row in reader:
rgz_to_atlas[row['ID_RGZ']] = row['ID']
all_subjects = data.get_all_subjects(survey='atlas', field=field)
if test:
all_subjects = all_subjects.limit(10)
for subject in all_subjects:
ra, dec = subject['coords']
zooniverse_id = subject['zooniverse_id']
rgz_source_id = subject['metadata']['source']
if rgz_source_id not in rgz_to_atlas:
logging.debug('Skipping %s; no matching ATLAS component.',
zooniverse_id)
continue
name = rgz_to_atlas[rgz_source_id]
# Store the results.
coords.append((ra, dec))
names.append(name)
zooniverse_ids.append(zooniverse_id)
elif (field == 'elais'):
atlascatalogue = ascii.read(config['data_sources']['atlas_catalogue'])
ras, decs = atlascatalogue['RA_deg'], atlascatalogue['Dec_deg']
e_ids = atlascatalogue['ID']
fields = atlascatalogue['field']
# Store the results.
for ra, dec, e_id, field_ in zip(ras, decs, e_ids, fields):
if (field_ == 'ELAIS-S1'):
coords.append((ra, dec))
names.append(e_id)
zooniverse_ids.append(e_id)
n_cdfs = len(names)
# Sort the data by Zooniverse ID.
coords_to_zooniverse_ids = dict(zip(coords, zooniverse_ids))
names_to_zooniverse_ids = dict(zip(names, zooniverse_ids))
coords.sort(key=coords_to_zooniverse_ids.get)
names.sort(key=names_to_zooniverse_ids.get)
zooniverse_ids.sort()
# Begin to store the data. We will have two tables: one for numeric data,
# and one for strings. We will have to preallocate the numeric table so that
# we aren't storing huge amounts of image data in memory.
# Strings.
dtype = [('zooniverse_id', '<S{}'.format(MAX_ZOONIVERSE_ID_LENGTH)),
('name', '<S{}'.format(MAX_NAME_LENGTH))]
string_data = numpy.array(list(zip(zooniverse_ids, names)), dtype=dtype)
cdfs.create_dataset('string', data=string_data, dtype=dtype)
# Numeric.
image_size = (config['surveys']['atlas']['fits_width'] *
config['surveys']['atlas']['fits_height'])
# RA, DEC, radio, (distance to SWIRE object added later)
dim = (n_cdfs, 1 + 1 + image_size)
numeric = cdfs.create_dataset('_numeric', shape=dim, dtype='float32')
# Load image patches and store numeric data.
with astropy.io.fits.open(
config['data_sources']['atlas_{}_image'.format(field)],
ignore_blank=True) as atlas_image:
wcs = astropy.wcs.WCS(atlas_image[0].header).dropaxis(3).dropaxis(2)
pix_coords = wcs.all_world2pix(coords, FITS_CONVENTION)
assert pix_coords.shape[1] == 2
logging.debug('Fetching %d ATLAS images.', len(pix_coords))
for index, (x, y) in enumerate(pix_coords):
radio = atlas_image[0].data[
0, 0, # stokes, freq
int(y) - config['surveys']['atlas']['fits_height'] // 2:
int(y) + config['surveys']['atlas']['fits_height'] // 2,
int(x) - config['surveys']['atlas']['fits_width'] // 2:
int(x) + config['surveys']['atlas']['fits_width'] // 2]
numeric[index, 0] = coords[index][0]
numeric[index, 1] = coords[index][1]
numeric[index, 2:2 + image_size] = radio.reshape(-1)
logging.debug('ATLAS imported.')
|
Python
|
def import_norris(f_h5):
"""Imports the Norris et al. (2006) labels.
f_h5: crowdastro HDF5 file with WISE or SWIRE already imported.
"""
ir_survey = f_h5.attrs['ir_survey']
ir_positions = f_h5['/{}/cdfs/numeric'.format(ir_survey)][:, :2]
ir_tree = sklearn.neighbors.KDTree(ir_positions)
norris_dat = astropy.io.ascii.read(config['data_sources']['norris_coords'])
norris_swire = norris_dat['SWIRE']
norris_coords = []
for s in norris_swire:
s = s.strip()
if len(s) < 19:
continue
# e.g. J032931.44-281722.0
ra_hr = s[1:3]
ra_min = s[3:5]
ra_sec = s[5:10]
dec_sgn = s[10]
dec_deg = s[11:13]
dec_min = s[13:15]
dec_sec = s[15:19]
ra = '{} {} {}'.format(ra_hr, ra_min, ra_sec)
dec = '{}{} {} {}'.format(dec_sgn, dec_deg, dec_min, dec_sec)
logging.debug('Reading Norris coordinate: {}; {}'.format(ra, dec))
coord = SkyCoord(ra=ra, dec=dec,
unit=('hourangle, deg'))
norris_coords.append(coord)
norris_labels = numpy.zeros((ir_positions.shape[0],))
for skycoord in norris_coords:
# Find a neighbour.
ra = skycoord.ra.degree
dec = skycoord.dec.degree
((dist,),), ((ir,),) = ir_tree.query([(ra, dec)])
if dist < config['surveys'][ir_survey]['distance_cutoff']:
norris_labels[ir] = 1
f_h5.create_dataset('/{}/cdfs/norris_labels'.format(ir_survey),
data=norris_labels)
|
def import_norris(f_h5):
"""Imports the Norris et al. (2006) labels.
f_h5: crowdastro HDF5 file with WISE or SWIRE already imported.
"""
ir_survey = f_h5.attrs['ir_survey']
ir_positions = f_h5['/{}/cdfs/numeric'.format(ir_survey)][:, :2]
ir_tree = sklearn.neighbors.KDTree(ir_positions)
norris_dat = astropy.io.ascii.read(config['data_sources']['norris_coords'])
norris_swire = norris_dat['SWIRE']
norris_coords = []
for s in norris_swire:
s = s.strip()
if len(s) < 19:
continue
# e.g. J032931.44-281722.0
ra_hr = s[1:3]
ra_min = s[3:5]
ra_sec = s[5:10]
dec_sgn = s[10]
dec_deg = s[11:13]
dec_min = s[13:15]
dec_sec = s[15:19]
ra = '{} {} {}'.format(ra_hr, ra_min, ra_sec)
dec = '{}{} {} {}'.format(dec_sgn, dec_deg, dec_min, dec_sec)
logging.debug('Reading Norris coordinate: {}; {}'.format(ra, dec))
coord = SkyCoord(ra=ra, dec=dec,
unit=('hourangle, deg'))
norris_coords.append(coord)
norris_labels = numpy.zeros((ir_positions.shape[0],))
for skycoord in norris_coords:
# Find a neighbour.
ra = skycoord.ra.degree
dec = skycoord.dec.degree
((dist,),), ((ir,),) = ir_tree.query([(ra, dec)])
if dist < config['surveys'][ir_survey]['distance_cutoff']:
norris_labels[ir] = 1
f_h5.create_dataset('/{}/cdfs/norris_labels'.format(ir_survey),
data=norris_labels)
|
Python
|
def import_fan(f_h5):
"""Imports the Fan et al. (2015) labels.
f_h5: crowdastro HDF5 file with WISE or SWIRE already imported.
"""
ir_survey = f_h5.attrs['ir_survey']
ir_names = f_h5['/{}/cdfs/string'.format(ir_survey)]
ir_positions = f_h5['/{}/cdfs/numeric'.format(ir_survey)][:, :2]
ir_tree = sklearn.neighbors.KDTree(ir_positions)
fan_coords = []
with open(config['data_sources']['fan_swire'], 'r') as fan_dat:
for row in csv.DictReader(fan_dat):
ra_hr = row['swire'][8:10]
ra_min = row['swire'][10:12]
ra_sec = row['swire'][12:17]
dec_sgn = row['swire'][17]
dec_deg = row['swire'][18:20]
dec_min = row['swire'][20:22]
dec_sec = row['swire'][22:26]
ra = '{} {} {}'.format(ra_hr, ra_min, ra_sec)
dec = '{}{} {} {}'.format(dec_sgn, dec_deg, dec_min, dec_sec)
fan_coords.append((ra, dec))
fan_labels = numpy.zeros((ir_positions.shape[0],))
for ra, dec in fan_coords:
# Find a neighbour.
skycoord = SkyCoord(ra=ra, dec=dec, unit=('hourangle', 'deg'))
ra = skycoord.ra.degree
dec = skycoord.dec.degree
((dist,),), ((ir,),) = ir_tree.query([(ra, dec)])
if dist < config['surveys'][ir_survey]['distance_cutoff']:
fan_labels[ir] = 1
f_h5.create_dataset('/{}/cdfs/fan_labels'.format(ir_survey),
data=fan_labels)
|
def import_fan(f_h5):
"""Imports the Fan et al. (2015) labels.
f_h5: crowdastro HDF5 file with WISE or SWIRE already imported.
"""
ir_survey = f_h5.attrs['ir_survey']
ir_names = f_h5['/{}/cdfs/string'.format(ir_survey)]
ir_positions = f_h5['/{}/cdfs/numeric'.format(ir_survey)][:, :2]
ir_tree = sklearn.neighbors.KDTree(ir_positions)
fan_coords = []
with open(config['data_sources']['fan_swire'], 'r') as fan_dat:
for row in csv.DictReader(fan_dat):
ra_hr = row['swire'][8:10]
ra_min = row['swire'][10:12]
ra_sec = row['swire'][12:17]
dec_sgn = row['swire'][17]
dec_deg = row['swire'][18:20]
dec_min = row['swire'][20:22]
dec_sec = row['swire'][22:26]
ra = '{} {} {}'.format(ra_hr, ra_min, ra_sec)
dec = '{}{} {} {}'.format(dec_sgn, dec_deg, dec_min, dec_sec)
fan_coords.append((ra, dec))
fan_labels = numpy.zeros((ir_positions.shape[0],))
for ra, dec in fan_coords:
# Find a neighbour.
skycoord = SkyCoord(ra=ra, dec=dec, unit=('hourangle', 'deg'))
ra = skycoord.ra.degree
dec = skycoord.dec.degree
((dist,),), ((ir,),) = ir_tree.query([(ra, dec)])
if dist < config['surveys'][ir_survey]['distance_cutoff']:
fan_labels[ir] = 1
f_h5.create_dataset('/{}/cdfs/fan_labels'.format(ir_survey),
data=fan_labels)
|
Python
|
def contains(bbox, point):
"""Checks if point is within bbox.
bbox: [[x0, x1], [y0, y1]]
point: [x, y]
-> bool
"""
return (bbox[0][0] <= point[0] <= bbox[0][1] and
bbox[1][0] <= point[1] <= bbox[1][1])
|
def contains(bbox, point):
"""Checks if point is within bbox.
bbox: [[x0, x1], [y0, y1]]
point: [x, y]
-> bool
"""
return (bbox[0][0] <= point[0] <= bbox[0][1] and
bbox[1][0] <= point[1] <= bbox[1][1])
|
Python
|
def make_radio_combination_signature(radio_annotation, wcs, atlas_positions,
subject, pix_offset):
"""Generates a unique signature for a radio annotation.
radio_annotation: 'radio' dictionary from a classification.
wcs: World coordinate system associated with the ATLAS image.
atlas_positions: [[RA, DEC]] NumPy array.
subject: RGZ subject dict.
pix_offset: (x, y) pixel position of this radio subject on the ATLAS image.
-> Something immutable
"""
from . import rgz_data as data
# TODO(MatthewJA): This only works on ATLAS. Generalise.
# My choice of immutable object will be stringified crowdastro ATLAS
# indices.
zooniverse_id = subject['zooniverse_id']
subject_fits = data.get_radio_fits(subject)
subject_wcs = astropy.wcs.WCS(subject_fits.header)
atlas_ids = []
x_offset, y_offset = pix_offset
for c in radio_annotation.values():
# Note that the x scale is not the same as the IR scale, but the scale
# factor is included in the annotation, so I have multiplied this out
# here for consistency.
scale_width = c.get('scale_width', '')
scale_height = c.get('scale_height', '')
if scale_width:
scale_width = float(scale_width)
else:
# Sometimes, there's no scale, so I've included a default scale.
scale_width = config['surveys']['atlas']['scale_width']
if scale_height:
scale_height = float(scale_height)
else:
scale_height = config['surveys']['atlas']['scale_height']
# These numbers are in terms of the PNG images, so I need to multiply by
# the click-to-fits ratio.
scale_width *= config['surveys']['atlas']['click_to_fits_x']
scale_height *= config['surveys']['atlas']['click_to_fits_y']
subject_bbox = [
[
float(c['xmin']) * scale_width,
float(c['xmax']) * scale_width,
],
[
float(c['ymin']) * scale_height,
float(c['ymax']) * scale_height,
],
]
# ...and by the mosaic ratio. There's probably double-up here, but this
# makes more sense.
scale_width *= config['surveys']['atlas']['mosaic_scale_x']
scale_height *= config['surveys']['atlas']['mosaic_scale_y']
# Get the bounding box of the radio source in pixels.
# Format: [xs, ys]
bbox = [
[
float(c['xmin']) * scale_width,
float(c['xmax']) * scale_width,
],
[
float(c['ymin']) * scale_height,
float(c['ymax']) * scale_height,
],
]
assert bbox[0][0] < bbox[0][1]
assert bbox[1][0] < bbox[1][1]
# Convert the bounding box into RA/DEC.
bbox = wcs.all_pix2world(bbox[0] + x_offset, bbox[1] + y_offset,
FITS_CONVENTION)
subject_bbox = subject_wcs.all_pix2world(subject_bbox[0],
subject_bbox[1], FITS_CONVENTION)
# TODO(MatthewJA): Remove (or disable) this sanity check.
# The bbox is backwards along the x-axis for some reason.
bbox[0] = bbox[0][::-1]
assert bbox[0][0] < bbox[0][1]
assert bbox[1][0] < bbox[1][1]
bbox = numpy.array(bbox)
# What is this radio source called? Check if we have an object in the
# bounding box. We'll cache these results because there is a lot of
# overlap.
cache_key = tuple(tuple(b) for b in bbox)
if cache_key in bbox_cache_:
index = bbox_cache_[cache_key]
else:
x_gt_min = atlas_positions[:, 0] >= bbox[0, 0]
x_lt_max = atlas_positions[:, 0] <= bbox[0, 1]
y_gt_min = atlas_positions[:, 1] >= bbox[1, 0]
y_lt_max = atlas_positions[:, 1] <= bbox[1, 1]
within = numpy.all([x_gt_min, x_lt_max, y_gt_min, y_lt_max], axis=0)
indices = numpy.where(within)[0]
if len(indices) == 0:
logging.debug('Skipping radio source not in catalogue for '
'%s', zooniverse_id)
continue
else:
if len(indices) > 1:
logging.debug('Found multiple (%d) ATLAS matches '
'for %s', len(indices), zooniverse_id)
index = indices[0]
bbox_cache_[cache_key] = index
atlas_ids.append(str(index))
atlas_ids.sort()
if not atlas_ids:
raise CatalogueError('No catalogued radio sources.')
return ';'.join(atlas_ids)
|
def make_radio_combination_signature(radio_annotation, wcs, atlas_positions,
subject, pix_offset):
"""Generates a unique signature for a radio annotation.
radio_annotation: 'radio' dictionary from a classification.
wcs: World coordinate system associated with the ATLAS image.
atlas_positions: [[RA, DEC]] NumPy array.
subject: RGZ subject dict.
pix_offset: (x, y) pixel position of this radio subject on the ATLAS image.
-> Something immutable
"""
from . import rgz_data as data
# TODO(MatthewJA): This only works on ATLAS. Generalise.
# My choice of immutable object will be stringified crowdastro ATLAS
# indices.
zooniverse_id = subject['zooniverse_id']
subject_fits = data.get_radio_fits(subject)
subject_wcs = astropy.wcs.WCS(subject_fits.header)
atlas_ids = []
x_offset, y_offset = pix_offset
for c in radio_annotation.values():
# Note that the x scale is not the same as the IR scale, but the scale
# factor is included in the annotation, so I have multiplied this out
# here for consistency.
scale_width = c.get('scale_width', '')
scale_height = c.get('scale_height', '')
if scale_width:
scale_width = float(scale_width)
else:
# Sometimes, there's no scale, so I've included a default scale.
scale_width = config['surveys']['atlas']['scale_width']
if scale_height:
scale_height = float(scale_height)
else:
scale_height = config['surveys']['atlas']['scale_height']
# These numbers are in terms of the PNG images, so I need to multiply by
# the click-to-fits ratio.
scale_width *= config['surveys']['atlas']['click_to_fits_x']
scale_height *= config['surveys']['atlas']['click_to_fits_y']
subject_bbox = [
[
float(c['xmin']) * scale_width,
float(c['xmax']) * scale_width,
],
[
float(c['ymin']) * scale_height,
float(c['ymax']) * scale_height,
],
]
# ...and by the mosaic ratio. There's probably double-up here, but this
# makes more sense.
scale_width *= config['surveys']['atlas']['mosaic_scale_x']
scale_height *= config['surveys']['atlas']['mosaic_scale_y']
# Get the bounding box of the radio source in pixels.
# Format: [xs, ys]
bbox = [
[
float(c['xmin']) * scale_width,
float(c['xmax']) * scale_width,
],
[
float(c['ymin']) * scale_height,
float(c['ymax']) * scale_height,
],
]
assert bbox[0][0] < bbox[0][1]
assert bbox[1][0] < bbox[1][1]
# Convert the bounding box into RA/DEC.
bbox = wcs.all_pix2world(bbox[0] + x_offset, bbox[1] + y_offset,
FITS_CONVENTION)
subject_bbox = subject_wcs.all_pix2world(subject_bbox[0],
subject_bbox[1], FITS_CONVENTION)
# TODO(MatthewJA): Remove (or disable) this sanity check.
# The bbox is backwards along the x-axis for some reason.
bbox[0] = bbox[0][::-1]
assert bbox[0][0] < bbox[0][1]
assert bbox[1][0] < bbox[1][1]
bbox = numpy.array(bbox)
# What is this radio source called? Check if we have an object in the
# bounding box. We'll cache these results because there is a lot of
# overlap.
cache_key = tuple(tuple(b) for b in bbox)
if cache_key in bbox_cache_:
index = bbox_cache_[cache_key]
else:
x_gt_min = atlas_positions[:, 0] >= bbox[0, 0]
x_lt_max = atlas_positions[:, 0] <= bbox[0, 1]
y_gt_min = atlas_positions[:, 1] >= bbox[1, 0]
y_lt_max = atlas_positions[:, 1] <= bbox[1, 1]
within = numpy.all([x_gt_min, x_lt_max, y_gt_min, y_lt_max], axis=0)
indices = numpy.where(within)[0]
if len(indices) == 0:
logging.debug('Skipping radio source not in catalogue for '
'%s', zooniverse_id)
continue
else:
if len(indices) > 1:
logging.debug('Found multiple (%d) ATLAS matches '
'for %s', len(indices), zooniverse_id)
index = indices[0]
bbox_cache_[cache_key] = index
atlas_ids.append(str(index))
atlas_ids.sort()
if not atlas_ids:
raise CatalogueError('No catalogued radio sources.')
return ';'.join(atlas_ids)
|
Python
|
def parse_classification(classification, subject, atlas_positions, wcs,
pix_offset):
"""Converts a raw RGZ classification into a classification dict.
Scales all positions and flips y axis of clicks.
classification: RGZ classification dict.
subject: Associated RGZ subject dict.
atlas_positions: [[RA, DEC]] NumPy array.
wcs: World coordinate system of the ATLAS image.
pix_offset: (x, y) pixel position of this radio subject on the ATLAS image.
-> dict mapping radio signature to list of corresponding IR host pixel
locations.
"""
result = {}
n_invalid = 0
for annotation in classification['annotations']:
if 'radio' not in annotation:
# This is a metadata annotation and we can ignore it.
continue
if annotation['radio'] == 'No Contours':
# I'm not sure how this occurs. I'm going to ignore it.
continue
try:
radio_signature = make_radio_combination_signature(
annotation['radio'], wcs, atlas_positions,
subject, pix_offset)
except CatalogueError:
# Ignore invalid annotations.
n_invalid += 1
logging.debug('Ignoring invalid annotation for %s.',
subject['zooniverse_id'])
continue
ir_locations = []
if annotation['ir'] != 'No Sources':
for ir_click in annotation['ir']:
ir_x = float(annotation['ir'][ir_click]['x'])
ir_y = float(annotation['ir'][ir_click]['y'])
# Rescale to a consistent size.
ir_x *= config['surveys']['atlas']['click_to_fits_x']
ir_y *= config['surveys']['atlas']['click_to_fits_y']
# Ignore out-of-range data.
if not 0 <= ir_x <= config['surveys']['atlas']['fits_width']:
n_invalid += 1
continue
if not 0 <= ir_y <= config['surveys']['atlas']['fits_height']:
n_invalid += 1
continue
# Flip the y axis to match other data conventions.
ir_y = config['surveys']['atlas']['fits_height'] - ir_y
# Rescale to match the mosaic WCS.
ir_x *= config['surveys']['atlas']['mosaic_scale_x']
ir_y *= config['surveys']['atlas']['mosaic_scale_y']
# Move to the reference location of the radio subject.
ir_x += pix_offset[0]
ir_y += pix_offset[1]
# Convert the location into RA/DEC.
(ir_x,), (ir_y,) = wcs.wcs_pix2world([ir_x], [ir_y], 1)
ir_location = (ir_x, ir_y)
ir_locations.append(ir_location)
result[radio_signature] = ir_locations
if n_invalid:
logging.debug('%d invalid annotations for %s.', n_invalid,
subject['zooniverse_id'])
return result
|
def parse_classification(classification, subject, atlas_positions, wcs,
pix_offset):
"""Converts a raw RGZ classification into a classification dict.
Scales all positions and flips y axis of clicks.
classification: RGZ classification dict.
subject: Associated RGZ subject dict.
atlas_positions: [[RA, DEC]] NumPy array.
wcs: World coordinate system of the ATLAS image.
pix_offset: (x, y) pixel position of this radio subject on the ATLAS image.
-> dict mapping radio signature to list of corresponding IR host pixel
locations.
"""
result = {}
n_invalid = 0
for annotation in classification['annotations']:
if 'radio' not in annotation:
# This is a metadata annotation and we can ignore it.
continue
if annotation['radio'] == 'No Contours':
# I'm not sure how this occurs. I'm going to ignore it.
continue
try:
radio_signature = make_radio_combination_signature(
annotation['radio'], wcs, atlas_positions,
subject, pix_offset)
except CatalogueError:
# Ignore invalid annotations.
n_invalid += 1
logging.debug('Ignoring invalid annotation for %s.',
subject['zooniverse_id'])
continue
ir_locations = []
if annotation['ir'] != 'No Sources':
for ir_click in annotation['ir']:
ir_x = float(annotation['ir'][ir_click]['x'])
ir_y = float(annotation['ir'][ir_click]['y'])
# Rescale to a consistent size.
ir_x *= config['surveys']['atlas']['click_to_fits_x']
ir_y *= config['surveys']['atlas']['click_to_fits_y']
# Ignore out-of-range data.
if not 0 <= ir_x <= config['surveys']['atlas']['fits_width']:
n_invalid += 1
continue
if not 0 <= ir_y <= config['surveys']['atlas']['fits_height']:
n_invalid += 1
continue
# Flip the y axis to match other data conventions.
ir_y = config['surveys']['atlas']['fits_height'] - ir_y
# Rescale to match the mosaic WCS.
ir_x *= config['surveys']['atlas']['mosaic_scale_x']
ir_y *= config['surveys']['atlas']['mosaic_scale_y']
# Move to the reference location of the radio subject.
ir_x += pix_offset[0]
ir_y += pix_offset[1]
# Convert the location into RA/DEC.
(ir_x,), (ir_y,) = wcs.wcs_pix2world([ir_x], [ir_y], 1)
ir_location = (ir_x, ir_y)
ir_locations.append(ir_location)
result[radio_signature] = ir_locations
if n_invalid:
logging.debug('%d invalid annotations for %s.', n_invalid,
subject['zooniverse_id'])
return result
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.