def refgenie_init(genome_config_path, genome_server=DEFAULT_SERVER):
"""
Initialize a genome config file.
:param str genome_config_path: path to genome configuration file to
create/initialize
:param st genome_server: URL for a server
"""
# Set up default
rgc = RefGenConf(
OrderedDict({
CFG_FOLDER_KEY:
os.path.dirname(os.path.abspath(genome_config_path)),
CFG_SERVER_KEY:
genome_server,
CFG_GENOMES_KEY:
None
}))
_LOGGER.debug("RGC: {}".format(rgc))
if genome_config_path and not os.path.exists(genome_config_path):
rgc.write(genome_config_path)
_LOGGER.info("Wrote new refgenie genome configuration file: {}".format(
genome_config_path))
else:
_LOGGER.warning(
"Can't initialize, file exists: {} ".format(genome_config_path))
def test_genome_folder_is_pwd_if_no_folder_key_and_raw_entries_passed(
self, ro_rgc):
data = PathExAttMap(
{k: v
for k, v in ro_rgc.items() if k != CFG_FOLDER_KEY})
new_rgc = RefGenConf(entries=data)
assert os.getcwd() == new_rgc[CFG_FOLDER_KEY]
def main():
global rgc, _LOGGER
parser = build_parser()
args = parser.parse_args()
if not args.command:
parser.print_help()
print("No subcommand given")
sys.exit(1)
logger_args = (dict(name=PKG_NAME, fmt=LOG_FORMAT, level=5)
if args.debug else dict(name=PKG_NAME, fmt=LOG_FORMAT))
_LOGGER = logmuse.setup_logger(**logger_args)
selected_cfg = select_genome_config(args.config)
assert (
selected_cfg is not None
), "You must provide a config file or set the {} environment variable".format(
"or ".join(CFG_ENV_VARS))
# this RefGenConf object will be used in the server, so it's read-only
rgc = RefGenConf(filepath=selected_cfg, writable=False)
if args.command == "archive":
arp = ([parse_registry_path(x) for x in args.asset_registry_paths]
if args.asset_registry_paths is not None else None)
archive(rgc, arp, args.force, args.remove, selected_cfg,
args.genomes_desc)
elif args.command == "serve":
# the router imports need to be after the RefGenConf object is declared
with rgc as r:
purge_nonservable(r)
from .routers import private, version1, version2, version3
app.include_router(version3.router)
app.include_router(version1.router, prefix="/v1")
app.include_router(version2.router, prefix="/v2")
app.include_router(version3.router, prefix="/v3")
app.include_router(private.router, prefix=f"/{PRIVATE_API}")
uvicorn.run(app, host="0.0.0.0", port=args.port)
def test_init_success(self):
rgc = RefGenConf()
dirpath = tempfile.mkdtemp(prefix="/tmp/")
cfg_file_path = os.path.join(dirpath, "test.yaml")
rgc.initialize_config_file(filepath=cfg_file_path)
assert os.path.exists(cfg_file_path)
shutil.rmtree(dirpath)
def test_force_overwrite_asset(self, cfg_file, gname, aname, tname):
rgc = RefGenConf(filepath=cfg_file)
path = rgc.seek(genome_name=gname,
asset_name="fasta",
tag_name=tname,
enclosing_dir=True)
assert rgc.add(path, gname, aname, tname, force=True)
assert rgc.add(path, gname, aname, tname, force=True)
def test_pull_asset_works_with_nonwritable_and_writable_rgc(
cfg_file, gname, aname, tname, state):
rgc = RefGenConf(filepath=cfg_file, writable=state)
remove_asset_and_file(rgc, gname, aname, tname)
with mock.patch("refgenconf.refgenconf.query_yes_no", return_value=True):
print("\nPulling; genome: {}, asset: {}, tag: {}\n".format(
gname, aname, tname))
rgc.pull_asset(gname, aname, tname)
def test_pull_asset_updates_genome_config(cfg_file, gname, aname, tname):
"""
Test that the object that was identical prior to the asset pull differs afterwards
and the pulled asset metadata has been written to the config file
"""
ori_rgc = RefGenConf(filepath=cfg_file, writable=False)
rgc = RefGenConf(filepath=cfg_file, writable=False)
remove_asset_and_file(rgc, gname, aname, tname)
remove_asset_and_file(ori_rgc, gname, aname, tname)
# ori_rgc.remove_assets(gname, aname, tname)
assert ori_rgc.to_dict() == rgc.to_dict()
with mock.patch("refgenconf.refgenconf.query_yes_no", return_value=True):
print("\nPulling; genome: {}, asset: {}, tag: {}\n".format(gname, aname, tname))
rgc.pull(gname, aname, tname)
assert not ori_rgc.to_dict() == rgc.to_dict()
post_rgc = RefGenConf(filepath=cfg_file, writable=False)
assert isinstance(post_rgc.seek(gname, aname, tname), str)
def test_prelist_plugins_called(self, cfg_file):
with mock.patch("refgenconf.refgenconf.RefGenConf.plugins",
new_callable=mock.PropertyMock) as mock_plugins:
mock_plugins.return_value = PLUGINS_DICT
rgc = RefGenConf(cfg_file, writable=False)
rgc.list()
assert get_flag_pth(rgc)
os.remove(get_flag_pth(rgc))
assert not os.path.exists(get_flag_pth(rgc))
def test_correct_namespaces(self, namespaces, genome, cfg_file):
namespaces["pipeline"]["var_templates"]["refgenie_config"] = cfg_file
ret = looper_refgenie_populate(namespaces=namespaces)
assert "refgenie" in ret
rgc = RefGenConf(filepath=cfg_file)
assert all([
asset in ret["refgenie"][genome].keys()
for asset in rgc.list_assets_by_genome(genome=genome)
])
def test_seekr(self, remote_class, servers, reset, genome, asset):
unbound_rgc = RefGenConf()
unbound_rgc.subscribe(servers, no_write=True, reset=reset)
isinstance(
unbound_rgc.seekr(genome_name=genome,
asset_name=asset,
remote_class=remote_class),
str,
)
def test_cant_add_without_digest_set_first(self, cfg_file, gname, aname,
tname):
rgc = RefGenConf(filepath=cfg_file)
path = rgc.seek(genome_name=gname,
asset_name="fasta",
tag_name=tname,
enclosing_dir=True)
gname = gname + "_new"
assert not rgc.add(path, gname, aname, tname)
def test_path_overrides(self, namespaces, genome, cfg_file):
rgc = RefGenConf(filepath=cfg_file)
test_asset = rgc.list_assets_by_genome(genome=genome)[0]
namespaces["pipeline"]["var_templates"]["refgenie_config"] = cfg_file
namespaces["project"]["refgenie"]["path_overrides"][0][
"registry_path"] = f"{genome}/{test_asset}"
ret = looper_refgenie_populate(namespaces=namespaces)
assert "refgenie" in ret
assert ret["refgenie"][genome][test_asset][test_asset] == "REPLACEMENT"
def test_illegal_genomes_mapping_type_gets_converted_to_empty_mapping(
self, genomes, tmpdir):
rgc = RefGenConf(
entries={
CFG_FOLDER_KEY: tmpdir.strpath,
CFG_GENOMES_KEY: genomes,
CFG_SERVERS_KEY: DEFAULT_SERVER
})
res = rgc[CFG_GENOMES_KEY]
assert isinstance(res, PathExAttMap)
assert 0 == len(res)
def test_nofile(self, cfg_file, gname, aname, tname):
rgc = RefGenConf(filepath=cfg_file)
path = rgc.seek(genome_name=gname,
asset_name="fasta",
tag_name=tname,
enclosing_dir=True)
assert rgc.add(path,
gname,
aname,
tname,
seek_keys={"file": "b"},
force=True)
def test_listr(self, servers, reset, genome):
unbound_rgc = RefGenConf()
unbound_rgc.subscribe(servers, no_write=True, reset=reset)
remote_list = unbound_rgc.listr(genome=genome)
assert len(remote_list) == len(unbound_rgc[CFG_SERVERS_KEY])
if genome is not None:
assert all([
len(assets_dict.keys()) == 1 and genome in assets_dict.keys()
for url, assets_dict in remote_list.items()
])
else:
assert isinstance(remote_list, dict)
def test_list_remote(rgc, tmpdir):
""" Verify expected behavior of remote genome/asset listing. """
new_rgc = RefGenConf(
entries={
CFG_FOLDER_KEY: tmpdir.strpath,
CFG_SERVERS_KEY: DEFAULT_SERVER,
CFG_GENOMES_KEY: rgc[CFG_GENOMES_KEY]
})
new_rgc[CFG_SERVERS_KEY] = "http://staging.refgenomes.databio.org"
print("NEW RGC KEYS: {}".format(list(new_rgc.keys())))
with mock.patch("refgenconf.refgenconf._read_remote_data",
return_value=rgc.genomes):
genomes, assets = new_rgc.list_remote()
_assert_eq_as_sets(rgc.genomes_str(), genomes)
def test_list_remote(rgc, tmpdir):
""" Verify expected behavior of remote genome/asset listing. """
new_rgc = RefGenConf(
entries={
CFG_FOLDER_KEY: tmpdir.strpath,
CFG_SERVERS_KEY: [DEFAULT_SERVER],
CFG_GENOMES_KEY: rgc[CFG_GENOMES_KEY]
})
result = new_rgc.listr()
assert list(result.keys())[0].startswith(DEFAULT_SERVER)
for server_url, asset_dict in result.items():
assert isinstance(asset_dict, OrderedDict)
assert len(asset_dict) == len(
_download_json(DEFAULT_SERVER + "/genomes"))
def test_populater(self, remote_class, servers, reset, genome, asset):
demo, nested_demo = get_demo_dicts(genome=genome,
asset=asset,
str_len=50)
unbound_rgc = RefGenConf()
unbound_rgc.subscribe(servers, no_write=True, reset=reset)
assert unbound_rgc.seekr(genome_name=genome,
asset_name=asset,
remote_class=remote_class) in str(
unbound_rgc.populater(
glob=demo, remote_class=remote_class))
assert unbound_rgc.seekr(genome_name=genome,
asset_name=asset,
remote_class=remote_class) in str(
unbound_rgc.populater(
glob=nested_demo,
remote_class=remote_class))
def test_genome_folder_is_value_from_config_file_if_key_present(
self, tmpdir_factory, tmpdir, made_genome_config_file):
conf_file = tmpdir_factory.mktemp("data2").join(
"refgenie.yaml").strpath
expected = tmpdir.strpath
with open(made_genome_config_file, 'r') as fin, open(conf_file,
'w') as fout:
found = False
for l in fin:
if l.startswith(CFG_FOLDER_KEY):
fout.write("{}: {}\n".format(CFG_FOLDER_KEY, expected))
else:
fout.write(l)
if l.startswith(CFG_SERVERS_KEY):
found = True
if not found:
fout.write("{}: {}".format(CFG_SERVERS_KEY, DEFAULT_SERVER))
rgc = RefGenConf(filepath=conf_file)
assert expected != os.path.dirname(conf_file)
assert expected == rgc[CFG_FOLDER_KEY]
def test_all_server_local_mix(self, cfg_file_old):
"""
Test config upgrade from v0.3 to v0.4 when a mix of genomes in terms of
remote digest availability is in defined the old config
"""
old_rgc = _RefGenConfV03(cfg_file_old)
# get some old asset data on disk
g, a, t = "human_alu", "fasta", "default"
try:
pth = old_rgc.seek(g, "fasta", "default", strict_exists=True)
except MissingGenomeError:
src_url = f"http://big.databio.org/refgenie_raw/files.{g}.{a}.{a}"
target_archive = f"/tmp/old/{g}.fa.gz"
target_file = f"/tmp/old/{g}.fa"
target_dir = f"/tmp/old/{g}/{a}/{t}"
os.makedirs(target_dir, exist_ok=True)
urllib.request.urlretrieve(src_url, target_archive)
from subprocess import run
run(
f"gunzip {target_archive}; "
f"mv {target_file} {target_dir}",
shell=True,
)
old_rgc.add(
path=target_dir,
genome=g,
asset=a,
tag="default",
seek_keys={a: f"{g}.fa"},
force=True,
)
else:
print(f"{pth} exists")
finally:
upgrade_config(filepath=cfg_file_old,
target_version="0.4",
force=True)
rgc = RefGenConf(cfg_file_old)
assert rgc[CFG_VERSION_KEY] == REQ_CFG_VERSION
def ro_rgc(cfg_file):
return RefGenConf(filepath=cfg_file, writable=False)
def my_rgc(cfg_file):
return RefGenConf(filepath=cfg_file)
def rgc(made_genome_config_file):
"""Provide test case with a genome config instance."""
with open(made_genome_config_file, "r") as f:
return RefGenConf(entries=yaml.load(f, yaml.SafeLoader))
def test_preexisting_asset_prompt(self, cfg_file, gname, aname, tname):
rgc = RefGenConf(filepath=cfg_file)
path = rgc.seek(genome_name=gname, asset_name=aname, tag_name=tname)
with mock.patch("refgenconf.refgenconf.query_yes_no",
return_value=False):
assert not rgc.add(path, gname, aname, tname)
def test_nonexistent_file(self, cfg_file, pth, gname, aname, tname):
rgc = RefGenConf(filepath=cfg_file)
with pytest.raises(OSError):
rgc.add(pth, gname, aname, tname)
def main():
""" Primary workflow """
parser = logmuse.add_logging_options(build_argparser())
args, remaining_args = parser.parse_known_args()
global _LOGGER
_LOGGER = logmuse.logger_via_cli(args)
logmuse.logger_via_cli(args, name=refgenconf.__name__)
_LOGGER.debug("Args: {}".format(args))
if not args.command:
parser.print_help()
_LOGGER.error("No command given")
sys.exit(1)
gencfg = yacman.select_config(args.genome_config,
CFG_ENV_VARS,
check_exist=not args.command == INIT_CMD,
on_missing=lambda fp: fp)
if gencfg is None:
raise MissingGenomeConfigError(args.genome_config)
_LOGGER.debug("Determined genome config: {}".format(gencfg))
if args.command == INIT_CMD:
_LOGGER.info("Initializing refgenie genome configuration")
_writeable(os.path.dirname(gencfg), strict_exists=True)
refgenie_init(gencfg, args.genome_server)
sys.exit(0)
rgc = RefGenConf(gencfg)
if args.command == BUILD_CMD:
refgenie_build(rgc, args)
elif args.command == GET_ASSET_CMD:
_LOGGER.debug("getting asset: '{}/{}'".format(args.genome, args.asset))
print(" ".join(
[rgc.get_asset(args.genome, asset) for asset in args.asset]))
return
elif args.command == INSERT_CMD:
if len(args.asset) > 1:
raise NotImplementedError("Can only add 1 asset at a time")
else:
# recast from list to str
args.asset = args.asset[0]
refgenie_add(rgc, args)
elif args.command == PULL_CMD:
outdir = rgc[CFG_FOLDER_KEY]
if not os.path.exists(outdir):
raise MissingFolderError(outdir)
target = _key_to_name(CFG_FOLDER_KEY)
if not perm_check_x(outdir, target):
return
if not _single_folder_writeable(outdir):
_LOGGER.error("Insufficient permissions to write to {}: "
"{}".format(target, outdir))
return
rgc.pull_asset(args.genome,
args.asset,
gencfg,
unpack=not args.no_untar)
elif args.command in [LIST_LOCAL_CMD, LIST_REMOTE_CMD]:
pfx, genomes, assets = _exec_list(rgc, args.command == LIST_REMOTE_CMD)
_LOGGER.info("{} genomes: {}".format(pfx, genomes))
_LOGGER.info("{} assets:\n{}".format(pfx, assets))
def refgenie_build(gencfg, genome, asset_list, recipe_name, args):
"""
Runs the refgenie build recipe.
:param str gencfg: path to the genome configuration file
:param argparse.Namespace args: parsed command-line options/arguments
"""
rgc = RefGenConf(
filepath=gencfg,
writable=False,
skip_read_lock=_skip_lock(args.skip_read_lock, gencfg),
)
specified_args = _parse_user_build_input(args.files)
specified_params = _parse_user_build_input(args.params)
def _read_json_file(filepath):
"""
Read a JSON file
:param str filepath: path to the file to read
:return dict: read data
"""
with open(filepath, "r") as f:
data = json.load(f)
return data
if recipe_name and os.path.isfile(recipe_name) and recipe_name.endswith(
".json"):
recipe_name = _read_json_file(filepath=recipe_name)
if not hasattr(args, "outfolder") or not args.outfolder:
# Default to genome_folder
_LOGGER.debug("No outfolder provided, using genome config.")
args.outfolder = rgc.data_dir
def _build_asset(
genome,
asset_key,
tag,
build_pkg,
genome_outfolder,
specific_args,
specific_params,
alias,
**kwargs,
):
"""
Builds assets with pypiper and updates a genome config file.
This function actually run the build commands in a given build package,
and then update the refgenie config file.
:param str genome: The assembly key; e.g. 'mm10'.
:param str asset_key: The unique asset identifier; e.g. 'bowtie2_index'
:param dict build_pkg: A dict (see examples) specifying lists
of required input_assets, commands to run, and outputs to register as
assets.
"""
log_outfolder = os.path.abspath(
os.path.join(genome_outfolder, asset_key, tag, BUILD_STATS_DIR))
_LOGGER.info("Saving outputs to:\n- content: {}\n- logs: {}".format(
genome_outfolder, log_outfolder))
if args.docker:
# Set up some docker stuff
if args.volumes:
# TODO: is volumes list defined here?
volumes = volumes.append(genome_outfolder)
else:
volumes = genome_outfolder
if not _writeable(genome_outfolder):
_LOGGER.error(
"Insufficient permissions to write to output folder: {}".
format(genome_outfolder))
return
pm = pypiper.PipelineManager(name="refgenie",
outfolder=log_outfolder,
args=args)
tk = pypiper.NGSTk(pm=pm)
if args.docker:
pm.get_container(build_pkg[CONT], volumes)
_LOGGER.debug("Asset build package: " + str(build_pkg))
# create a bundle list to simplify calls below
gat = [genome, asset_key, tag]
# collect variables required to populate the command templates
asset_vars = get_asset_vars(
genome,
asset_key,
tag,
genome_outfolder,
specific_args,
specific_params,
**kwargs,
)
# populate command templates
# prior to populating, remove any seek_key parts from the keys, since these are not supported by format method
command_list_populated = [
x.format(**{k.split(".")[0]: v
for k, v in asset_vars.items()})
for x in build_pkg[CMD_LST]
]
# create output directory
tk.make_dir(asset_vars["asset_outfolder"])
target = os.path.join(log_outfolder,
TEMPLATE_TARGET.format(genome, asset_key, tag))
# add target command
command_list_populated.append("touch {target}".format(target=target))
_LOGGER.debug("Command populated: '{}'".format(
" ".join(command_list_populated)))
try:
# run build command
signal.signal(signal.SIGINT, _handle_sigint(gat))
pm.run(command_list_populated, target, container=pm.container)
except pypiper.exceptions.SubprocessError:
_LOGGER.error("asset '{}' build failed".format(asset_key))
return False
else:
# save build recipe to the JSON-formatted file
recipe_file_name = TEMPLATE_RECIPE_JSON.format(asset_key, tag)
with open(os.path.join(log_outfolder, recipe_file_name),
"w") as outfile:
json.dump(build_pkg, outfile)
# since the assets are always built to a standard dir structure, we
# can just stitch a path together for asset digest calculation
asset_dir = os.path.join(rgc.data_dir, *gat)
if not os.path.exists(asset_dir):
raise OSError("Could not compute asset digest. Path does not "
"exist: {}".format(asset_dir))
digest = get_dir_digest(asset_dir)
_LOGGER.info("Asset digest: {}".format(digest))
# add updates to config file
with rgc as r:
if asset_key == "fasta":
r.update_genomes(genome,
data={CFG_ALIASES_KEY: [alias]},
force_digest=genome)
r.update_assets(
*gat[0:2],
data={CFG_ASSET_DESC_KEY: build_pkg[DESC]},
force_digest=genome,
)
r.update_tags(
*gat,
force_digest=genome,
data={
CFG_ASSET_PATH_KEY: asset_key,
CFG_ASSET_CHECKSUM_KEY: digest,
},
)
r.update_seek_keys(
*gat,
force_digest=genome,
keys={
k: v.format(**asset_vars)
for k, v in build_pkg[ASSETS].items()
},
)
r.set_default_pointer(*gat, force_digest=genome)
pm.stop_pipeline()
return True
for a in asset_list:
asset_key = a["asset"]
asset_tag = a["tag"] or rgc.get_default_tag(
genome, a["asset"], use_existing=False)
recipe_name = recipe_name or asset_key
if isinstance(
recipe_name,
dict) or (isinstance(recipe_name, str)
and recipe_name in asset_build_packages.keys()):
if isinstance(recipe_name, dict):
_LOGGER.info("Using custom recipe: \n{}".format(recipe_name))
asset_build_package = _check_recipe(recipe_name)
recipe_name = asset_build_package["name"]
else:
asset_build_package = _check_recipe(
asset_build_packages[recipe_name])
# handle user-requested parents for the required assets
input_assets = {}
parent_assets = []
specified_asset_keys, specified_assets = None, None
if args.assets is not None:
parsed_parents_input = _parse_user_build_input(args.assets)
specified_asset_keys = list(parsed_parents_input.keys())
specified_assets = list(parsed_parents_input.values())
_LOGGER.debug(f"Custom assets requested: {args.assets}")
if not specified_asset_keys and isinstance(args.assets, list):
_LOGGER.warning(
"Specified parent assets format is invalid. Using defaults."
)
for req_asset in asset_build_package[REQ_ASSETS]:
req_asset_data = parse_registry_path(req_asset[KEY])
# for each req asset see if non-default parents were requested
if (specified_asset_keys is not None
and req_asset_data["asset"] in specified_asset_keys):
parent_data = parse_registry_path(
specified_assets[specified_asset_keys.index(
req_asset_data["asset"])])
g, a, t, s = (
parent_data["genome"],
parent_data["asset"],
parent_data["tag"]
or rgc.get_default_tag(genome, parent_data["asset"]),
parent_data["seek_key"],
)
else: # if no custom parents requested for the req asset, use default one
default = parse_registry_path(req_asset[DEFAULT])
g, a, t, s = (
genome,
default["asset"],
rgc.get_default_tag(genome, default["asset"]),
req_asset_data["seek_key"],
)
parent_assets.append("{}/{}:{}".format(
rgc.get_genome_alias_digest(g, fallback=True), a, t))
input_assets[req_asset[KEY]] = _seek(rgc, g, a, t, s)
_LOGGER.debug("Using parents: {}".format(", ".join(parent_assets)))
_LOGGER.debug("Provided files: {}".format(specified_args))
_LOGGER.debug("Provided parameters: {}".format(specified_params))
for required_file in asset_build_package[REQ_FILES]:
if (specified_args is None
or required_file[KEY] not in specified_args.keys()):
raise ValueError(
"Path to the '{x}' input ({desc}) is required, but not provided. "
"Specify it with: --files {x}=/path/to/{x}_file".
format(x=required_file[KEY], desc=required_file[DESC]))
for required_param in asset_build_package[REQ_PARAMS]:
if specified_params is None:
specified_params = {}
if required_param[KEY] not in specified_params.keys():
if required_param[DEFAULT] is None:
raise ValueError(
"Value for the parameter '{x}' ({desc}) is required, but not provided. "
"Specify it with: --params {x}=value".format(
x=required_param[KEY],
desc=required_param[DESC]))
else:
specified_params.update(
{required_param[KEY]: required_param[DEFAULT]})
_LOGGER.info("Building '{}/{}:{}' using '{}' recipe".format(
genome, asset_key, asset_tag, recipe_name))
ori_genome = genome
if recipe_name == "fasta":
if (genome in rgc.genomes_list()
and "fasta" in rgc.list_assets_by_genome(genome)):
pretag = rgc.get_default_tag(genome, "fasta")
_LOGGER.warning(
"'{g}' genome is already initialized with other fasta asset ({g}/{a}:{t})"
.format(g=genome, a=asset_key, t=pretag))
genome = rgc.get_genome_alias_digest(alias=genome,
fallback=True)
else:
# if the recipe is "fasta" we first initialiaze the genome, based on the provided path to the input FASTA file
genome, _ = rgc.initialize_genome(
fasta_path=specified_args["fasta"],
alias=ori_genome,
skip_alias_write=True,
)
else:
try:
genome = rgc.get_genome_alias_digest(genome, fallback=True)
except UndefinedAliasError:
_LOGGER.error("Genome '{}' has not been initialized yet; "
"no key found for this alias".format(genome))
return
recipe_name = None
genome_outfolder = os.path.join(args.outfolder, genome)
if not _build_asset(
genome,
asset_key,
asset_tag,
asset_build_package,
genome_outfolder,
specified_args,
specified_params,
ori_genome,
**input_assets,
):
log_path = os.path.abspath(
os.path.join(
genome_outfolder,
asset_key,
asset_tag,
BUILD_STATS_DIR,
ORI_LOG_NAME,
))
_LOGGER.info(
"'{}/{}:{}' was not added to the config, but directory has been left in place. "
"See the log file for details: {}".format(
genome, asset_key, asset_tag, log_path))
return
_LOGGER.info("Finished building '{}' asset".format(asset_key))
with rgc as r:
# update asset relationships
r.update_relatives_assets(genome, asset_key, asset_tag,
parent_assets) # adds parents
for i in parent_assets:
parsed_parent = parse_registry_path(i)
# adds child (currently built asset) to the parent
r.update_relatives_assets(
parsed_parent["genome"],
parsed_parent["asset"],
parsed_parent["tag"],
["{}/{}:{}".format(genome, asset_key, asset_tag)],
True,
)
if args.genome_description is not None:
_LOGGER.debug(
"adding genome ({}) description: '{}'".format(
genome, args.genome_description))
r.update_genomes(
genome, {CFG_GENOME_DESC_KEY: args.genome_description})
if args.tag_description is not None:
_LOGGER.debug(
"adding tag ({}/{}:{}) description: '{}'".format(
genome, asset_key, asset_tag,
args.tag_description))
r.update_tags(
genome,
asset_key,
asset_tag,
{CFG_TAG_DESC_KEY: args.tag_description},
)
rgc._symlink_alias(genome, asset_key, asset_tag)
else:
_raise_missing_recipe_error(recipe_name)
def main():
""" Primary workflow """
parser = logmuse.add_logging_options(build_argparser())
args, remaining_args = parser.parse_known_args()
global _LOGGER
_LOGGER = logmuse.logger_via_cli(args, make_root=True)
_LOGGER.debug("refgenie {}".format(__version__))
_LOGGER.debug("Args: {}".format(args))
if not args.command:
parser.print_help()
_LOGGER.error("No command given")
sys.exit(1)
gencfg = refgenconf.select_genome_config(
filename=args.genome_config,
check_exist=not args.command == INIT_CMD,
on_missing=lambda fp: fp,
strict_env=True)
if gencfg is None:
raise MissingGenomeConfigError(args.genome_config)
_LOGGER.debug("Determined genome config: {}".format(gencfg))
# From user input we want to construct a list of asset dicts, where each
# asset has a genome name, asset name, and tag
if "asset_registry_paths" in args and args.asset_registry_paths:
_LOGGER.debug("Found registry_path: {}".format(
args.asset_registry_paths))
asset_list = [
parse_registry_path(x) for x in args.asset_registry_paths
]
for a in asset_list:
# every asset must have a genome, either provided via registry path
# or the args.genome arg.
if not a["genome"]:
if args.genome:
a["genome"] = args.genome
else:
_LOGGER.error(
"Provided asset registry path ({}/{}:{}) is invalid. See help for usage reference."
.format(a["genome"], a["asset"], a["tag"]))
sys.exit(1)
else:
if args.genome and args.genome != a["genome"]:
_LOGGER.warn(
"Two different genomes specified for asset '{}'.".
format(a["asset"]))
else:
if args.command in GENOME_ONLY_REQUIRED and not args.genome:
parser.error("You must provide either a genome or a registry path")
sys.exit(1)
if args.command in ASSET_REQUIRED:
parser.error("You must provide an asset registry path")
sys.exit(1)
if args.command == INIT_CMD:
_LOGGER.debug("Initializing refgenie genome configuration")
rgc = RefGenConf(entries=OrderedDict(
{
CFG_VERSION_KEY: REQ_CFG_VERSION,
CFG_FOLDER_KEY: os.path.dirname(os.path.abspath(gencfg)),
CFG_SERVERS_KEY: args.genome_server or [DEFAULT_SERVER],
CFG_GENOMES_KEY: None
}))
rgc.initialize_config_file(os.path.abspath(gencfg))
elif args.command == BUILD_CMD:
if not all(
[x["genome"] == asset_list[0]["genome"] for x in asset_list]):
_LOGGER.error("Build can only build assets for one genome")
sys.exit(1)
recipe_name = None
if args.recipe:
if len(asset_list) > 1:
_LOGGER.error(
"Recipes cannot be specified for multi-asset builds")
sys.exit(1)
recipe_name = args.recipe
if args.requirements:
for a in asset_list:
recipe = recipe_name or a["asset"]
if recipe not in asset_build_packages.keys():
_raise_missing_recipe_error(recipe)
_LOGGER.info("'{}' recipe requirements: ".format(recipe))
_make_asset_build_reqs(recipe)
sys.exit(0)
refgenie_build(gencfg, asset_list[0]["genome"], asset_list,
recipe_name, args)
elif args.command == GET_ASSET_CMD:
rgc = RefGenConf(filepath=gencfg, writable=False)
check = args.check_exists if args.check_exists else None
for a in asset_list:
_LOGGER.debug("getting asset: '{}/{}.{}:{}'".format(
a["genome"], a["asset"], a["seek_key"], a["tag"]))
print(
rgc.seek(a["genome"],
a["asset"],
a["tag"],
a["seek_key"],
strict_exists=check))
return
elif args.command == INSERT_CMD:
rgc = RefGenConf(filepath=gencfg, writable=False)
if len(asset_list) > 1:
raise NotImplementedError("Can only add 1 asset at a time")
else:
refgenie_add(rgc, asset_list[0], args.path, args.force)
elif args.command == PULL_CMD:
rgc = RefGenConf(filepath=gencfg, writable=False)
force = None if not args.force else True
outdir = rgc[CFG_FOLDER_KEY]
if not os.path.exists(outdir):
raise MissingFolderError(outdir)
target = _key_to_name(CFG_FOLDER_KEY)
if not perm_check_x(outdir, target):
return
if not _single_folder_writeable(outdir):
_LOGGER.error("Insufficient permissions to write to {}: {}".format(
target, outdir))
return
for a in asset_list:
rgc.pull(a["genome"],
a["asset"],
a["tag"],
unpack=not args.no_untar,
force=force)
elif args.command in [LIST_LOCAL_CMD, LIST_REMOTE_CMD]:
rgc = RefGenConf(filepath=gencfg, writable=False)
if args.command == LIST_REMOTE_CMD:
num_servers = 0
# Keep all servers so that child updates maintain server list
server_list = rgc[CFG_SERVERS_KEY]
bad_servers = []
for server_url in rgc[CFG_SERVERS_KEY]:
num_servers += 1
try:
rgc[CFG_SERVERS_KEY] = server_url
pfx, genomes, assets, recipes = _exec_list(
rgc, args.command == LIST_REMOTE_CMD, args.genome)
if assets is None and genomes is None:
continue
_LOGGER.info("{} genomes: {}".format(pfx, genomes))
if args.command != LIST_REMOTE_CMD: # Not implemented yet
_LOGGER.info("{} recipes: {}".format(pfx, recipes))
_LOGGER.info("{} assets:\n{}\n".format(pfx, assets))
except (DownloadJsonError, ConnectionError):
bad_servers.append(server_url)
continue
if num_servers >= len(server_list) and bad_servers:
_LOGGER.error(
"Could not list assets from the following server(s): {}".
format(bad_servers))
# Restore original server list, even when we couldn't find assets on a server
rgc[CFG_SERVERS_KEY] = server_list
else: # Only check local assets once
_LOGGER.info("Server subscriptions: {}".format(", ".join(
rgc[CFG_SERVERS_KEY])))
pfx, genomes, assets, recipes = _exec_list(
rgc, args.command == LIST_REMOTE_CMD, args.genome)
_LOGGER.info("{} genomes: {}".format(pfx, genomes))
if args.command != LIST_REMOTE_CMD: # Not implemented yet
_LOGGER.info("{} recipes: {}".format(pfx, recipes))
_LOGGER.info("{} assets:\n{}".format(pfx, assets))
elif args.command == GETSEQ_CMD:
rgc = RefGenConf(filepath=gencfg, writable=False)
rgc.getseq(rgc, args.genome, args.locus)
elif args.command == REMOVE_CMD:
force = args.force
rgc = RefGenConf(filepath=gencfg)
for a in asset_list:
a["tag"] = a["tag"] or rgc.get_default_tag(
a["genome"], a["asset"], use_existing=False)
_LOGGER.debug("Determined tag for removal: {}".format(a["tag"]))
if a["seek_key"] is not None:
raise NotImplementedError(
"You can't remove a specific seek_key.")
bundle = [a["genome"], a["asset"], a["tag"]]
try:
if not rgc.is_asset_complete(*bundle):
with rgc as r:
r.cfg_remove_assets(*bundle)
_LOGGER.info(
"Removed an incomplete asset '{}/{}:{}'".format(
*bundle))
return
except (KeyError, MissingAssetError, MissingGenomeError):
_LOGGER.info("Asset '{}/{}:{}' does not exist".format(*bundle))
return
if len(asset_list) > 1:
if not query_yes_no(
"Are you sure you want to remove {} assets?".format(
len(asset_list))):
_LOGGER.info("Action aborted by the user")
return
force = True
for a in asset_list:
rgc.remove(genome=a["genome"],
asset=a["asset"],
tag=a["tag"],
force=force)
elif args.command == TAG_CMD:
rgc = RefGenConf(filepath=gencfg)
if len(asset_list) > 1:
raise NotImplementedError("Can only tag 1 asset at a time")
if args.default:
# set the default tag and exit
with rgc as r:
r.set_default_pointer(a["genome"], a["asset"], a["tag"], True)
sys.exit(0)
rgc.tag(a["genome"], a["asset"], a["tag"], args.tag)
elif args.command == ID_CMD:
rgc = RefGenConf(filepath=gencfg, writable=False)
if len(asset_list) == 1:
g, a = asset_list[0]["genome"], asset_list[0]["asset"]
t = asset_list[0]["tag"] or rgc.get_default_tag(g, a)
print(rgc.id(g, a, t))
return
for asset in asset_list:
g, a = asset["genome"], asset["asset"]
t = asset["tag"] or rgc.get_default_tag(g, a)
print("{}/{}:{},".format(g, a, t) + rgc.id(g, a, t))
return
elif args.command == SUBSCRIBE_CMD:
rgc = RefGenConf(filepath=gencfg, writable=False)
rgc.subscribe(urls=args.genome_server, reset=args.reset)
return
elif args.command == UNSUBSCRIBE_CMD:
rgc = RefGenConf(filepath=gencfg, writable=False)
rgc.unsubscribe(urls=args.genome_server)
return
def refgenie_build(gencfg, genome, asset_list, recipe_name, args):
"""
Runs the refgenie build recipe.
:param str gencfg: path to the genome configuration file
:param argparse.Namespace args: parsed command-line options/arguments
"""
rgc = RefGenConf(filepath=gencfg, writable=False)
specified_args = _parse_user_build_input(args.files)
specified_params = _parse_user_build_input(args.params)
if not hasattr(args, "outfolder") or not args.outfolder:
# Default to genome_folder
_LOGGER.debug("No outfolder provided, using genome config.")
args.outfolder = rgc[CFG_FOLDER_KEY]
_LOGGER.debug("Default config file: {}".format(default_config_file()))
if args.config_file and not os.path.isfile(args.config_file):
_LOGGER.debug("Config file path isn't a file: {}".format(
args.config_file))
args.config_file = default_config_file()
def build_asset(genome, asset_key, tag, build_pkg, genome_outfolder,
specific_args, specific_params, **kwargs):
"""
Builds assets with pypiper and updates a genome config file.
This function actually run the build commands in a given build package,
and then update the refgenie config file.
:param str genome: The assembly key; e.g. 'mm10'.
:param str asset_key: The unique asset identifier; e.g. 'bowtie2_index'
:param dict build_pkg: A dict (see examples) specifying lists
of required input_assets, commands to run, and outputs to register as
assets.
"""
log_outfolder = os.path.abspath(
os.path.join(genome_outfolder, asset_key, tag, BUILD_STATS_DIR))
_LOGGER.info("Saving outputs to:\n- content: {}\n- logs: {}".format(
genome_outfolder, log_outfolder))
if args.docker:
# Set up some docker stuff
if args.volumes:
# TODO: is volumes list defined here?
volumes = volumes.append(genome_outfolder)
else:
volumes = genome_outfolder
if not _writeable(genome_outfolder):
_LOGGER.error(
"Insufficient permissions to write to output folder: {}".
format(genome_outfolder))
return
pm = pypiper.PipelineManager(name="refgenie",
outfolder=log_outfolder,
args=args)
tk = pypiper.NGSTk(pm=pm)
if args.docker:
pm.get_container(build_pkg[CONT], volumes)
_LOGGER.debug("Asset build package: " + str(build_pkg))
gat = [genome, asset_key,
tag] # create a bundle list to simplify calls below
# collect variables required to populate the command templates
asset_vars = get_asset_vars(genome, asset_key, tag, genome_outfolder,
specific_args, specific_params, **kwargs)
# populate command templates
# prior to populating, remove any seek_key parts from the keys, since these are not supported by format method
command_list_populated = [
x.format(**{k.split(".")[0]: v
for k, v in asset_vars.items()})
for x in build_pkg[CMD_LST]
]
# create output directory
tk.make_dir(asset_vars["asset_outfolder"])
target = os.path.join(log_outfolder,
TEMPLATE_TARGET.format(genome, asset_key, tag))
# add target command
command_list_populated.append("touch {target}".format(target=target))
_LOGGER.debug("Command populated: '{}'".format(
" ".join(command_list_populated)))
try:
# run build command
signal.signal(signal.SIGINT, _handle_sigint(gat))
pm.run(command_list_populated, target, container=pm.container)
except pypiper.exceptions.SubprocessError:
_LOGGER.error("asset '{}' build failed".format(asset_key))
return False
else:
# save build recipe to the JSON-formatted file
recipe_file_name = TEMPLATE_RECIPE_JSON.format(asset_key, tag)
with open(os.path.join(log_outfolder, recipe_file_name),
'w') as outfile:
json.dump(build_pkg, outfile)
# update and write refgenie genome configuration
with rgc as r:
r.update_assets(*gat[0:2],
data={CFG_ASSET_DESC_KEY: build_pkg[DESC]})
r.update_tags(*gat, data={CFG_ASSET_PATH_KEY: asset_key})
r.update_seek_keys(*gat,
keys={
k: v.format(**asset_vars)
for k, v in build_pkg[ASSETS].items()
})
# in order to conveniently get the path to digest we update the tags metadata in two steps
digest = get_dir_digest(
r.get_asset(genome, asset_key, tag, enclosing_dir=True),
pm)
r.update_tags(*gat, data={CFG_ASSET_CHECKSUM_KEY: digest})
_LOGGER.info("Asset digest: {}".format(digest))
r.set_default_pointer(*gat)
pm.stop_pipeline()
return True
for a in asset_list:
asset_key = a["asset"]
asset_tag = a["tag"] or rgc.get_default_tag(
genome, a["asset"], use_existing=False)
recipe_name = recipe_name or asset_key
if recipe_name in asset_build_packages.keys():
asset_build_package = _check_recipe(
asset_build_packages[recipe_name])
# handle user-requested parents for the required assets
input_assets = {}
parent_assets = []
specified_asset_keys, specified_assets = None, None
if args.assets is not None:
parsed_parents_input = _parse_user_build_input(args.assets)
specified_asset_keys, specified_assets = \
list(parsed_parents_input.keys()), list(parsed_parents_input.values())
_LOGGER.debug("Custom assets requested: {}".format(
args.assets))
if not specified_asset_keys and isinstance(args.assets, list):
_LOGGER.warning(
"Specified parent assets format is invalid. Using defaults."
)
for req_asset in asset_build_package[REQ_ASSETS]:
req_asset_data = parse_registry_path(req_asset[KEY])
# for each req asset see if non-default parents were requested
if specified_asset_keys is not None and req_asset_data[
"asset"] in specified_asset_keys:
parent_data = \
parse_registry_path(specified_assets[specified_asset_keys.index(req_asset_data["asset"])])
g, a, t, s = parent_data["genome"], \
parent_data["asset"], \
parent_data["tag"] or rgc.get_default_tag(genome, parent_data["asset"]), \
parent_data["seek_key"]
else: # if no custom parents requested for the req asset, use default one
default = parse_registry_path(req_asset[DEFAULT])
g, a, t, s = genome, default["asset"], \
rgc.get_default_tag(genome, default["asset"]), \
req_asset_data["seek_key"]
parent_assets.append("{}/{}:{}".format(g, a, t))
input_assets[req_asset[KEY]] = _seek(rgc, g, a, t, s)
_LOGGER.debug("Using parents: {}".format(", ".join(parent_assets)))
_LOGGER.debug("Provided files: {}".format(specified_args))
_LOGGER.debug("Provided parameters: {}".format(specified_params))
for required_file in asset_build_package[REQ_FILES]:
if specified_args is None or required_file[
KEY] not in specified_args.keys():
raise ValueError(
"Path to the '{x}' input ({desc}) is required, but not provided. "
"Specify it with: --files {x}=/path/to/{x}_file".
format(x=required_file[KEY], desc=required_file[DESC]))
for required_param in asset_build_package[REQ_PARAMS]:
if specified_params is None:
specified_params = {}
if required_param[KEY] not in specified_params.keys():
if required_param[DEFAULT] is None:
raise ValueError(
"Value for the parameter '{x}' ({desc}) is required, but not provided. "
"Specify it with: --params {x}=value".format(
x=required_param[KEY],
desc=required_param[DESC]))
else:
specified_params.update(
{required_param[KEY]: required_param[DEFAULT]})
genome_outfolder = os.path.join(args.outfolder, genome)
_LOGGER.info("Building '{}/{}:{}' using '{}' recipe".format(
genome, asset_key, asset_tag, recipe_name))
if recipe_name == 'fasta' and genome in rgc.genomes_list() \
and 'fasta' in rgc.list_assets_by_genome(genome):
_LOGGER.warning(
"'{g}' genome is already initialized with other fasta asset ({g}/{a}:{t}). "
"It will be re-initialized.".format(g=genome,
a=asset_key,
t=asset_tag))
if not build_asset(genome, asset_key, asset_tag,
asset_build_package, genome_outfolder,
specified_args, specified_params, **
input_assets):
log_path = os.path.abspath(
os.path.join(genome_outfolder, asset_key, asset_tag,
BUILD_STATS_DIR, ORI_LOG_NAME))
_LOGGER.info(
"'{}/{}:{}' was not added to the config, but directory has been left in place. "
"See the log file for details: {}".format(
genome, asset_key, asset_tag, log_path))
return
# If the recipe was a fasta, we init the genome
if recipe_name == 'fasta':
_LOGGER.info("Computing initial genome digest...")
collection_checksum, content_checksums = \
fasta_checksum(_seek(rgc, genome, asset_key, asset_tag, "fasta"))
_LOGGER.info("Initializing genome...")
refgenie_initg(rgc, genome, content_checksums)
_LOGGER.info("Finished building '{}' asset".format(asset_key))
with rgc as r:
# update asset relationships
r.update_relatives_assets(genome, asset_key, asset_tag,
parent_assets) # adds parents
for i in parent_assets:
parsed_parent = parse_registry_path(i)
# adds child (currently built asset) to the parent
r.update_relatives_assets(
parsed_parent["genome"], parsed_parent["asset"],
parsed_parent["tag"],
["{}/{}:{}".format(genome, asset_key, asset_tag)],
True)
if args.genome_description is not None:
_LOGGER.debug(
"adding genome ({}) description: '{}'".format(
genome, args.genome_description))
r.update_genomes(
genome, {CFG_GENOME_DESC_KEY: args.genome_description})
if args.tag_description is not None:
_LOGGER.debug(
"adding tag ({}/{}:{}) description: '{}'".format(
genome, asset_key, asset_tag,
args.tag_description))
r.update_tags(genome, asset_key, asset_tag,
{CFG_TAG_DESC_KEY: args.tag_description})
if recipe_name == "fasta":
# to save config lock time when building fasta assets
# (genome initialization takes some time for large genomes) we repeat the
# conditional here for writing the computed genome digest
r.update_genomes(
genome, data={CFG_CHECKSUM_KEY: collection_checksum})
else:
_raise_missing_recipe_error(recipe_name)
def test_invalid_path(self, pth):
rgc = RefGenConf()
with pytest.raises(TypeError):
rgc.initialize_config_file(filepath=pth)