wehub-resource-sync
407 lines
14 KiB
Python
407 lines
14 KiB
Python
"""For parsing code snippets and locating relevant dependencies/imports so that they could be installed to the containers' images"""
|
|
|
|
from contextlib import contextmanager
|
|
import os
|
|
import sys
|
|
import re
|
|
import logging
|
|
import ast
|
|
from docopt import docopt
|
|
import requests
|
|
from pathlib import Path
|
|
from yarg import json2package
|
|
from yarg.exceptions import HTTPError
|
|
|
|
REGEXP = [
|
|
re.compile(r'^import (.+)$'),
|
|
re.compile(r'^from ((?!\.+).*?) import (?:.*)$')
|
|
]
|
|
|
|
logging.disable(sys.maxsize)
|
|
PIPS_FOLDER_PATH = os.path.join(os.path.dirname(Path(os.path.realpath(__file__)).parent), "pips")
|
|
DOCKER_REQUIREMENTS_PATH = os.path.join(os.path.dirname(Path(os.path.realpath(__file__)).parent.parent), "docker/docker/requirements.txt")
|
|
|
|
@contextmanager
|
|
def _open(filename=None, mode='r'):
|
|
"""Open a file or ``sys.stdout`` depending on the provided filename.
|
|
|
|
Args:
|
|
filename (str): The path to the file that should be opened. If
|
|
``None`` or ``'-'``, ``sys.stdout`` or ``sys.stdin`` is
|
|
returned depending on the desired mode. Defaults to ``None``.
|
|
mode (str): The mode that should be used to open the file.
|
|
|
|
Yields:
|
|
A file handle.
|
|
|
|
"""
|
|
if not filename or filename == '-':
|
|
if not mode or 'r' in mode:
|
|
file = sys.stdin
|
|
elif 'w' in mode:
|
|
file = sys.stdout
|
|
else:
|
|
raise ValueError('Invalid mode for file: {}'.format(mode))
|
|
else:
|
|
file = open(filename, mode)
|
|
|
|
try:
|
|
yield file
|
|
finally:
|
|
if file not in (sys.stdin, sys.stdout):
|
|
file.close()
|
|
|
|
|
|
def get_all_imports(
|
|
contents):
|
|
imports = set()
|
|
raw_imports = set()
|
|
candidates = []
|
|
ignore_errors = False
|
|
|
|
tree = ast.parse(contents)
|
|
for node in ast.walk(tree):
|
|
if isinstance(node, ast.Import):
|
|
for subnode in node.names:
|
|
raw_imports.add(subnode.name)
|
|
elif isinstance(node, ast.ImportFrom):
|
|
raw_imports.add(node.module)
|
|
|
|
# Clean up imports
|
|
for name in [n for n in raw_imports if n]:
|
|
# Sanity check: Name could have been None if the import
|
|
# statement was as ``from . import X``
|
|
# Cleanup: We only want to first part of the import.
|
|
# Ex: from django.conf --> django.conf. But we only want django
|
|
# as an import.
|
|
cleaned_name, _, _ = name.partition('.')
|
|
imports.add(cleaned_name)
|
|
|
|
packages = imports - (set(candidates) & imports)
|
|
logging.debug('Found packages: {0}'.format(packages))
|
|
|
|
with open(os.path.join(PIPS_FOLDER_PATH, "stdlib.txt"), "r") as f:
|
|
data = {x.strip() for x in f}
|
|
|
|
return list(packages - data)
|
|
|
|
|
|
def generate_requirements_file(path, imports, symbol):
|
|
with _open(path, "w") as out_file:
|
|
logging.debug('Writing {num} requirements: {imports} to {file}'.format(
|
|
num=len(imports),
|
|
file=path,
|
|
imports=", ".join([x['name'] for x in imports])
|
|
))
|
|
fmt = '{name}' + symbol + '{version}'
|
|
out_file.write('\n'.join(
|
|
fmt.format(**item) if item['version'] else '{name}'.format(**item)
|
|
for item in imports) + '\n')
|
|
|
|
|
|
def get_imports_info(
|
|
imports, pypi_server="https://pypi.python.org/pypi/", proxy=None):
|
|
result = []
|
|
|
|
# transformers most times require torch and sentencepiece
|
|
if ('transformers' in imports):
|
|
imports.extend(['torch', 'sentencepiece'])
|
|
|
|
for item in imports:
|
|
try:
|
|
logging.warning(
|
|
'Import named "%s" not found locally. '
|
|
'Trying to resolve it at the PyPI server.',
|
|
item
|
|
)
|
|
response = requests.get(
|
|
"{0}{1}/json".format(pypi_server, item), proxies=proxy)
|
|
if response.status_code == 200:
|
|
if hasattr(response.content, 'decode'):
|
|
data = json2package(response.content.decode())
|
|
else:
|
|
data = json2package(response.content)
|
|
elif response.status_code >= 300:
|
|
raise HTTPError(status_code=response.status_code,
|
|
reason=response.reason)
|
|
except HTTPError:
|
|
logging.warning(
|
|
'Package "%s" does not exist or network problems', item)
|
|
continue
|
|
logging.warning(
|
|
'Import named "%s" was resolved to "%s:%s" package (%s).\n'
|
|
'Please, verify manually the final list of requirements.txt '
|
|
'to avoid possible dependency confusions.',
|
|
item,
|
|
data.name,
|
|
data.latest_release_id,
|
|
data.pypi_url
|
|
)
|
|
result.append({'name': item, 'version': data.latest_release_id})
|
|
return result
|
|
|
|
|
|
def get_locally_installed_packages(encoding=None):
|
|
packages = []
|
|
ignore = ["tests", "_tests", "egg", "EGG", "info"]
|
|
for path in sys.path:
|
|
for root, dirs, files in os.walk(path):
|
|
for item in files:
|
|
if "top_level" in item:
|
|
item = os.path.join(root, item)
|
|
with open(item, "r", encoding=encoding) as f:
|
|
package = root.split(os.sep)[-1].split("-")
|
|
try:
|
|
top_level_modules = f.read().strip().split("\n")
|
|
except: # NOQA
|
|
# TODO: What errors do we intend to suppress here?
|
|
continue
|
|
|
|
# filter off explicitly ignored top-level modules
|
|
# such as test, egg, etc.
|
|
filtered_top_level_modules = list()
|
|
|
|
for module in top_level_modules:
|
|
if (
|
|
(module not in ignore) and
|
|
(package[0] not in ignore)
|
|
):
|
|
# append exported top level modules to the list
|
|
filtered_top_level_modules.append(module)
|
|
|
|
version = None
|
|
if len(package) > 1:
|
|
version = package[1].replace(
|
|
".dist", "").replace(".egg", "")
|
|
|
|
# append package: top_level_modules pairs
|
|
# instead of top_level_module: package pairs
|
|
packages.append({
|
|
'name': package[0],
|
|
'version': version,
|
|
'exports': filtered_top_level_modules
|
|
})
|
|
return packages
|
|
|
|
|
|
def get_import_local(imports, encoding=None):
|
|
local = get_locally_installed_packages()
|
|
result = []
|
|
if ('transformers' in imports):
|
|
imports.extend(['torch', 'sentencepiece'])
|
|
for item in imports:
|
|
# search through local packages
|
|
for package in local:
|
|
# if candidate import name matches export name
|
|
# or candidate import name equals to the package name
|
|
# append it to the result
|
|
if item in package['exports'] or item == package['name']:
|
|
result.append(package)
|
|
|
|
# removing duplicates of package/version
|
|
# had to use second method instead of the previous one,
|
|
# because we have a list in the 'exports' field
|
|
# https://stackoverflow.com/questions/9427163/remove-duplicate-dict-in-list-in-python
|
|
name_set = set()
|
|
result_unique = []
|
|
for n, i in enumerate(result):
|
|
if i['name'] not in name_set:
|
|
result_unique.append(i)
|
|
name_set.add(i['name'])
|
|
|
|
#result_unique = [i for n, i in enumerate(result) if i not in result[n+1:]]
|
|
|
|
return result_unique
|
|
|
|
|
|
def get_pkg_names(pkgs):
|
|
"""Get PyPI package names from a list of imports.
|
|
|
|
Args:
|
|
pkgs (List[str]): List of import names.
|
|
|
|
Returns:
|
|
List[str]: The corresponding PyPI package names.
|
|
|
|
"""
|
|
result = set()
|
|
with open(os.path.join(PIPS_FOLDER_PATH, "mapping.txt"), "r") as f:
|
|
data = dict(x.strip().split(":") for x in f)
|
|
for pkg in pkgs:
|
|
# Look up the mapped requirement. If a mapping isn't found,
|
|
# simply use the package name.
|
|
result.add(data.get(pkg, pkg))
|
|
# Return a sorted list for backward compatibility.
|
|
return sorted(result, key=lambda s: s.lower())
|
|
|
|
|
|
def get_name_without_alias(name):
|
|
if "import " in name:
|
|
match = REGEXP[0].match(name.strip())
|
|
if match:
|
|
name = match.groups(0)[0]
|
|
return name.partition(' as ')[0].partition('.')[0].strip()
|
|
|
|
|
|
def join(f):
|
|
return os.path.join(os.path.dirname(__file__), f)
|
|
|
|
|
|
def parse_requirements(file_):
|
|
"""Parse a requirements formatted file.
|
|
|
|
Traverse a string until a delimiter is detected, then split at said
|
|
delimiter, get module name by element index, create a dict consisting of
|
|
module:version, and add dict to list of parsed modules.
|
|
|
|
Args:
|
|
file_: File to parse.
|
|
|
|
Raises:
|
|
OSerror: If there's any issues accessing the file.
|
|
|
|
Returns:
|
|
tuple: The contents of the file, excluding comments.
|
|
"""
|
|
modules = []
|
|
# For the dependency identifier specification, see
|
|
# https://www.python.org/dev/peps/pep-0508/#complete-grammar
|
|
delim = ["<", ">", "=", "!", "~"]
|
|
|
|
try:
|
|
f = open(file_, "r")
|
|
except OSError:
|
|
logging.error("Failed on file: {}".format(file_))
|
|
raise
|
|
else:
|
|
try:
|
|
data = [x.strip() for x in f.readlines() if x != "\n"]
|
|
finally:
|
|
f.close()
|
|
|
|
data = [x for x in data if x[0].isalpha()]
|
|
|
|
for x in data:
|
|
# Check for modules w/o a specifier.
|
|
if not any([y in x for y in delim]):
|
|
modules.append({"name": x, "version": None})
|
|
for y in x:
|
|
if y in delim:
|
|
module = x.split(y)
|
|
module_name = module[0]
|
|
module_version = module[-1].replace("=", "")
|
|
module = {"name": module_name, "version": module_version}
|
|
|
|
if module not in modules:
|
|
modules.append(module)
|
|
|
|
break
|
|
|
|
return modules
|
|
|
|
|
|
def compare_modules(file_, imports):
|
|
"""Compare modules in a file to imported modules in a project.
|
|
|
|
Args:
|
|
file_ (str): File to parse for modules to be compared.
|
|
imports (tuple): Modules being imported in the project.
|
|
|
|
Returns:
|
|
tuple: The modules not imported in the project, but do exist in the
|
|
specified file.
|
|
"""
|
|
modules = parse_requirements(file_)
|
|
|
|
imports = [imports[i]["name"] for i in range(len(imports))]
|
|
modules = [modules[i]["name"] for i in range(len(modules))]
|
|
modules_not_imported = set(modules) - set(imports)
|
|
|
|
return modules_not_imported
|
|
|
|
|
|
def diff(file_, imports):
|
|
"""Display the difference between modules in a file and imported modules.""" # NOQA
|
|
modules_not_imported = compare_modules(file_, imports)
|
|
|
|
logging.info(
|
|
"The following modules are in {} but do not seem to be imported: "
|
|
"{}".format(file_, ", ".join(x for x in modules_not_imported)))
|
|
|
|
|
|
def clean(file_, imports):
|
|
"""Remove modules that aren't imported in project from file."""
|
|
modules_not_imported = compare_modules(file_, imports)
|
|
|
|
if len(modules_not_imported) == 0:
|
|
logging.info("Nothing to clean in " + file_)
|
|
return
|
|
|
|
re_remove = re.compile("|".join(modules_not_imported))
|
|
to_write = []
|
|
|
|
try:
|
|
f = open(file_, "r+")
|
|
except OSError:
|
|
logging.error("Failed on file: {}".format(file_))
|
|
raise
|
|
else:
|
|
try:
|
|
for i in f.readlines():
|
|
if re_remove.match(i) is None:
|
|
to_write.append(i)
|
|
f.seek(0)
|
|
f.truncate()
|
|
|
|
for i in to_write:
|
|
f.write(i)
|
|
finally:
|
|
f.close()
|
|
|
|
logging.info("Successfully cleaned up requirements in " + file_)
|
|
|
|
|
|
def dynamic_versioning(scheme, imports):
|
|
"""Enables dynamic versioning with <compat>, <gt> or <non-pin> schemes."""
|
|
if scheme == "no-pin":
|
|
imports = [{"name": item["name"], "version": ""} for item in imports]
|
|
symbol = ""
|
|
elif scheme == "gt":
|
|
symbol = ">="
|
|
elif scheme == "compat":
|
|
symbol = "~="
|
|
return imports, symbol
|
|
|
|
|
|
def extract_dependencies(contents, args = {}, path = DOCKER_REQUIREMENTS_PATH):
|
|
candidates = get_all_imports(contents)
|
|
candidates = get_pkg_names(candidates)
|
|
logging.debug("Found imports: " + ", ".join(candidates))
|
|
pypi_server = "https://pypi.python.org/pypi/"
|
|
proxy = None
|
|
logging.debug("Getting packages information from Local/PyPI")
|
|
local = get_import_local(candidates, encoding=None)
|
|
|
|
#commonly used enough to be worth adding
|
|
candidates.append('pickle')
|
|
# check if candidate name is found in
|
|
# the list of exported modules, installed locally
|
|
# and the package name is not in the list of local module names
|
|
# it add to difference
|
|
difference = [x for x in candidates if
|
|
# aggregate all export lists into one
|
|
# flatten the list
|
|
# check if candidate is in exports
|
|
x.lower() not in [y for x in local for y in x['exports']]
|
|
and
|
|
# check if candidate is package names
|
|
x.lower() not in [x['name'] for x in local]]
|
|
imports = local + get_imports_info(difference,
|
|
proxy=proxy,
|
|
pypi_server=pypi_server)
|
|
# sort imports based on lowercase name of package, similar to `pip freeze`.
|
|
imports = sorted(imports, key=lambda x: x['name'].lower())
|
|
generate_requirements_file(path, imports, "==")
|
|
logging.info("Successfully saved requirements file in " + path)
|
|
return imports
|