logger = logging.getLogger("conda_forge_tick.auto_tick")
PR_LIMIT = 5
MAX_PR_LIMIT = 50
MIGRATORS: MutableSequence[Migrator] = [
Version(
pr_limit=PR_LIMIT * 2,
piggy_back_migrations=[
Jinja2VarsCleanup(),
PipMigrator(),
LicenseMigrator(),
CondaForgeYAMLCleanup(),
ExtraJinja2KeysCleanup(),
],
),
]
BOT_RERUN_LABEL = {
"name": "bot-rerun",
"color": "#191970",
"description": (
"Apply this label if you want the bot to retry "
"issuing a particular pull-request"
),
}
def run(
def add_rebuild_migration_yaml(
migrators: MutableSequence[Migrator],
gx: nx.DiGraph,
package_names: Sequence[str],
output_to_feedstock: Mapping[str, str],
excluded_feedstocks: MutableSet[str],
migration_yaml: str,
config: dict = {},
migration_name: str = "",
pr_limit: int = PR_LIMIT,
) -> None:
"""Adds rebuild migrator.
Parameters
----------
migrators : list of Migrator
The list of migrators to run.
gx : networkx.DiGraph
The feedstock graph
package_names : list of str
The package who's pin was moved
output_to_feedstock : dict of str
Mapping of output name to feedstock name
excluded_feedstocks : set of str
Feedstock names which should never be included in the migration
migration_yaml : str
The raw yaml for the migration variant dict
config: dict
The __migrator contents of the migration
migration_name: str
Name of the migration
pr_limit : int, optional
The number of PRs per hour, defaults to 5
"""
total_graph = create_rebuild_graph(
gx,
package_names,
excluded_feedstocks,
include_noarch=config.get("include_noarch", False),
)
# Note at this point the graph is made of all packages that have a
# dependency on the pinned package via Host, run, or test.
# Some packages don't have a host section so we use their
# build section in its place.
package_names = {
p if p in gx.nodes else output_to_feedstock[p] for p in package_names
} - excluded_feedstocks
top_level = {
node
for node in {gx.successors(package_name) for package_name in package_names}
if (node in total_graph) and len(list(total_graph.predecessors(node))) == 0
}
cycles = list(nx.simple_cycles(total_graph))
migrator = MigrationYaml(
migration_yaml,
graph=total_graph,
pr_limit=pr_limit,
name=migration_name,
top_level=top_level,
cycles=cycles,
piggy_back_migrations=[
Jinja2VarsCleanup(),
PipMigrator(),
LicenseMigrator(),
CondaForgeYAMLCleanup(),
ExtraJinja2KeysCleanup(),
],
**config,
)
print(f"bump number is {migrator.bump_number}")
migrators.append(migrator)
def add_rebuild_migration_yaml(
migrators: MutableSequence[Migrator],
gx: nx.DiGraph,
package_names: Sequence[str],
output_to_feedstock: Mapping[str, str],
excluded_feedstocks: MutableSet[str],
migration_yaml: str,
config: dict = {},
migration_name: str = "",
pr_limit: int = PR_LIMIT,
) -> None:
"""Adds rebuild migrator.
Parameters
----------
migrators : list of Migrator
The list of migrators to run.
gx : networkx.DiGraph
The feedstock graph
package_names : list of str
The package who's pin was moved
output_to_feedstock : dict of str
Mapping of output name to feedstock name
excluded_feedstocks : set of str
Feedstock names which should never be included in the migration
migration_yaml : str
The raw yaml for the migration variant dict
config: dict
The __migrator contents of the migration
migration_name: str
Name of the migration
pr_limit : int, optional
The number of PRs per hour, defaults to 5
"""
total_graph = copy.deepcopy(gx)
for node, node_attrs in gx.nodes.items():
attrs: "AttrsTypedDict" = node_attrs["payload"]
requirements = attrs.get("requirements", {})
host = requirements.get("host", set())
build = requirements.get("build", set())
bh = host or build
criteria = bh & set(package_names) and (
"noarch" not in attrs.get("meta_yaml", {}).get("build", {})
)
# get host/build, run and test and launder them through outputs
# this should fix outputs related issues (eg gdal)
rq = set(
map(
lambda x: gx.graph["outputs_lut"].get(x, x),
(host or build)
| requirements.get("run", set())
| requirements.get("test", set()),
)
)
for e in list(total_graph.in_edges(node)):
if e[0] not in rq:
total_graph.remove_edge(*e)
if not any([criteria]) or node in excluded_feedstocks:
pluck(total_graph, node)
# post plucking we can have several strange cases, lets remove all selfloops
total_graph.remove_edges_from(nx.selfloop_edges(total_graph))
package_names = {
p if p in gx.nodes else output_to_feedstock[p] for p in package_names
} - excluded_feedstocks
top_level = {
node
for node in {gx.successors(package_name) for package_name in package_names}
if (node in total_graph) and len(list(total_graph.predecessors(node))) == 0
}
cycles = list(nx.simple_cycles(total_graph))
migrator = MigrationYaml(
migration_yaml,
graph=total_graph,
pr_limit=pr_limit,
name=migration_name,
top_level=top_level,
cycles=cycles,
piggy_back_migrations=[
PipMigrator(),
LicenseMigrator(),
CondaForgeYAMLCleanup(),
ExtraJinja2KeysCleanup(),
],
**config,
)
print(f"bump number is {migrator.bump_number}")
migrators.append(migrator)
import os
from ruamel.yaml import YAML
from conda_forge_tick.migrators import (
Version,
ExtraJinja2KeysCleanup,
)
from test_migrators import run_test_migration
VERSION_CF = Version(piggy_back_migrations=[ExtraJinja2KeysCleanup()])
YAML_PATH = os.path.join(os.path.dirname(__file__), 'test_yaml')
def test_version_extra_jinja2_keys_cleanup(tmpdir):
with open(os.path.join(YAML_PATH, 'version_extra_jinja2_keys.yaml'),
'r') as fp:
in_yaml = fp.read()
with open(
os.path.join(YAML_PATH, 'version_extra_jinja2_keys_correct.yaml'),
'r',
) as fp:
out_yaml = fp.read()
os.makedirs(os.path.join(tmpdir, 'recipe'), exist_ok=True)
run_test_migration(
m=VERSION_CF,
inp=in_yaml,
output=out_yaml,
import os
from ruamel.yaml import YAML
from conda_forge_tick.migrators import (
Version,
ExtraJinja2KeysCleanup,
)
from test_migrators import run_test_migration
VERSION_CF = Version(
set(),
dict(),
dict(),
piggy_back_migrations=[ExtraJinja2KeysCleanup()],
)
YAML_PATH = os.path.join(os.path.dirname(__file__), "test_yaml")
def test_version_extra_jinja2_keys_cleanup(tmpdir):
with open(os.path.join(YAML_PATH, "version_extra_jinja2_keys.yaml"),
"r") as fp:
in_yaml = fp.read()
with open(
os.path.join(YAML_PATH, "version_extra_jinja2_keys_correct.yaml"),
"r",
) as fp:
out_yaml = fp.read()
def initialize_migrators(
github_username: str = "",
github_password: str = "",
github_token: Optional[str] = None,
dry_run: bool = False,
) -> Tuple[MigratorSessionContext, list, MutableSequence[Migrator]]:
temp = glob.glob("/tmp/*")
gx = load_graph()
smithy_version = eval_cmd("conda smithy --version").strip()
pinning_version = json.loads(
eval_cmd("conda list conda-forge-pinning --json"))[0]["version"]
migrators = []
add_arch_migrate(migrators, gx)
migration_factory(migrators, gx)
add_replacement_migrator(
migrators,
gx,
"matplotlib",
"matplotlib-base",
("Unless you need `pyqt`, recipes should depend only on "
"`matplotlib-base`."),
alt_migrator=MatplotlibBase,
)
create_migration_yaml_creator(migrators=migrators, gx=gx)
print("rebuild migration graph sizes:", flush=True)
for m in migrators:
print(
f' {getattr(m, "name", m)} graph size: '
f'{len(getattr(m, "graph", []))}',
flush=True,
)
print(" ", flush=True)
mctx = MigratorSessionContext(
circle_build_url=os.getenv("CIRCLE_BUILD_URL", ""),
graph=gx,
smithy_version=smithy_version,
pinning_version=pinning_version,
github_username=github_username,
github_password=github_password,
github_token=github_token,
dry_run=dry_run,
)
print("building package import maps and version migrator", flush=True)
python_nodes = {
n
for n, v in mctx.graph.nodes("payload")
if "python" in v.get("req", "")
}
python_nodes.update([
k for node_name, node in mctx.graph.nodes("payload")
for k in node.get("outputs_names", []) if node_name in python_nodes
], )
version_migrator = Version(
python_nodes=python_nodes,
pr_limit=PR_LIMIT * 4,
piggy_back_migrations=[
Jinja2VarsCleanup(),
DuplicateLinesCleanup(),
PipMigrator(),
LicenseMigrator(),
CondaForgeYAMLCleanup(),
ExtraJinja2KeysCleanup(),
Build2HostMigrator(),
NoCondaInspectMigrator(),
Cos7Config(),
],
)
migrators = [version_migrator] + migrators
print(" ", flush=True)
return mctx, temp, migrators
def main(args: "CLIArgs") -> None:
# start profiler
profile_profiler = cProfile.Profile()
profile_profiler.enable()
# logging
from .xonsh_utils import env
debug = env.get("CONDA_FORGE_TICK_DEBUG", False)
if debug:
setup_logger(logging.getLogger("conda_forge_tick"), level="debug")
else:
setup_logger(logging.getLogger("conda_forge_tick"))
github_username = env.get("USERNAME", "")
github_password = env.get("PASSWORD", "")
github_token = env.get("GITHUB_TOKEN")
global MIGRATORS
mctx, temp, MIGRATORS = initialize_migrators(
github_username=github_username,
github_password=github_password,
dry_run=args.dry_run,
github_token=github_token,
)
python_nodes = {
n
for n, v in mctx.graph.nodes("payload")
if "python" in v.get("req", "")
}
python_nodes.update([
k for node_name, node in mctx.graph.nodes("payload")
for k in node.get("outputs_names", []) if node_name in python_nodes
], )
imports_by_package, packages_by_import = create_package_import_maps(
python_nodes)
version_migrator = Version(
python_nodes=python_nodes,
imports_by_package=imports_by_package,
packages_by_import=packages_by_import,
pr_limit=PR_LIMIT * 2,
piggy_back_migrations=[
Jinja2VarsCleanup(),
PipMigrator(),
LicenseMigrator(),
CondaForgeYAMLCleanup(),
ExtraJinja2KeysCleanup(),
],
)
MIGRATORS = [version_migrator] + MIGRATORS
# compute the time per migrator
(num_nodes, time_per_migrator,
tot_time_per_migrator) = _compute_time_per_migrator(mctx, )
for i, migrator in enumerate(MIGRATORS):
if hasattr(migrator, "name"):
extra_name = "-%s" % migrator.name
else:
extra_name = ""
logger.info(
"Total migrations for %s%s: %d - gets %f seconds (%f percent)",
migrator.__class__.__name__,
extra_name,
num_nodes[i],
time_per_migrator[i],
time_per_migrator[i] / tot_time_per_migrator * 100,
)
for mg_ind, migrator in enumerate(MIGRATORS):
mmctx = MigratorContext(session=mctx, migrator=migrator)
migrator.bind_to_ctx(mmctx)
good_prs = 0
_mg_start = time.time()
effective_graph = mmctx.effective_graph
time_per = time_per_migrator[mg_ind]
if hasattr(migrator, "name"):
extra_name = "-%s" % migrator.name
else:
extra_name = ""
logger.info(
"Running migrations for %s%s: %d",
migrator.__class__.__name__,
extra_name,
len(effective_graph.nodes),
)
possible_nodes = list(migrator.order(effective_graph, mctx.graph))
# version debugging info
if isinstance(migrator, Version):
logger.info("possible version migrations:")
for node_name in possible_nodes:
with effective_graph.nodes[node_name]["payload"] as attrs:
logger.info(
" node|curr|new|attempts: %s|%s|%s|%d",
node_name,
attrs.get("version"),
attrs.get("new_version"),
(attrs.get("new_version_attempts", {}).get(
attrs.get("new_version", ""),
0,
)),
)
for node_name in possible_nodes:
with mctx.graph.nodes[node_name]["payload"] as attrs:
# Don't let CI timeout, break ahead of the timeout so we make certain
# to write to the repo
# TODO: convert these env vars
_now = time.time()
if ((_now - int(env.get("START_TIME", time.time())) > int(
env.get("TIMEOUT", 600)))
or good_prs >= migrator.pr_limit
or (_now - _mg_start) > time_per):
break
fctx = FeedstockContext(
package_name=node_name,
feedstock_name=attrs["feedstock_name"],
attrs=attrs,
)
print("\n", flush=True, end="")
logger.info(
"%s%s IS MIGRATING %s",
migrator.__class__.__name__.upper(),
extra_name,
fctx.package_name,
)
try:
# Don't bother running if we are at zero
if (args.dry_run or mctx.gh.rate_limit()["resources"]
["core"]["remaining"] == 0):
break
migrator_uid, pr_json = run(
feedstock_ctx=fctx,
migrator=migrator,
rerender=migrator.rerender,
protocol="https",
hash_type=attrs.get("hash_type", "sha256"),
)
# if migration successful
if migrator_uid:
d = frozen_to_json_friendly(migrator_uid)
# if we have the PR already do nothing
if d["data"] in [
existing_pr["data"]
for existing_pr in attrs.get("PRed", [])
]:
pass
else:
if not pr_json:
pr_json = {
"state": "closed",
"head": {
"ref": "<this_is_not_a_branch>"
},
}
d["PR"] = pr_json
attrs.setdefault("PRed", []).append(d)
attrs.update(
{
"smithy_version": mctx.smithy_version,
"pinning_version": mctx.pinning_version,
}, )
except github3.GitHubError as e:
if e.msg == "Repository was archived so is read-only.":
attrs["archived"] = True
else:
logger.critical(
"GITHUB ERROR ON FEEDSTOCK: %s",
fctx.feedstock_name,
)
if is_github_api_limit_reached(e, mctx.gh):
break
except URLError as e:
logger.exception("URLError ERROR")
attrs["bad"] = {
"exception": str(e),
"traceback": str(traceback.format_exc()).split("\n"),
"code": getattr(e, "code"),
"url": getattr(e, "url"),
}
except Exception as e:
logger.exception("NON GITHUB ERROR")
attrs["bad"] = {
"exception": str(e),
"traceback": str(traceback.format_exc()).split("\n"),
}
else:
if migrator_uid:
# On successful PR add to our counter
good_prs += 1
finally:
# Write graph partially through
if not args.dry_run:
dump_graph(mctx.graph)
eval_cmd(f"rm -rf {mctx.rever_dir}/*")
logger.info(os.getcwd())
for f in glob.glob("/tmp/*"):
if f not in temp:
eval_cmd(f"rm -rf {f}")
if not args.dry_run:
logger.info(
"API Calls Remaining: %d",
mctx.gh.rate_limit()["resources"]["core"]["remaining"],
)
logger.info("Done")
# stop profiler
profile_profiler.disable()
# human readable
s_stream = io.StringIO()
# TODO: There are other ways to do this, with more freedom
profile_stats = pstats.Stats(profile_profiler,
stream=s_stream).sort_stats("tottime", )
profile_stats.print_stats()
# get current time
now = datetime.now()
current_time = now.strftime("%d-%m-%Y") + "_" + now.strftime("%H_%M_%S")
# output to data
os.makedirs("profiler", exist_ok=True)
with open(f"profiler/{current_time}.txt", "w+") as f:
f.write(s_stream.getvalue())