add WgetPlugin with WgetExtractor, WarcExtractor, WgetBinary

This commit is contained in:
Nick Sweeting 2024-09-30 19:33:30 -07:00
parent 2a1645ba27
commit c4e040f11a
No known key found for this signature in database
2 changed files with 87 additions and 50 deletions

View file

@ -1,70 +1,106 @@
from typing import List
from abx.archivebox.base_plugin import BasePlugin, InstanceOf, BaseHook
import sys
from typing import List, Optional
from pathlib import Path
from rich import print
from pydantic import InstanceOf, Field, model_validator
from pydantic_pkgr import BinProvider, BinName
from abx.archivebox.base_plugin import BasePlugin, BaseHook
from abx.archivebox.base_configset import BaseConfigSet
from abx.archivebox.base_binary import BaseBinary, env, apt, brew
from abx.archivebox.base_extractor import BaseExtractor, ExtractorName
from archivebox.extractors.wget import wget_output_path
# class WgetToggleConfig(ConfigSet):
class WgetConfig(BaseConfigSet):
# SAVE_WGET: bool = True
# SAVE_WARC: bool = True
SAVE_WGET: bool = True
SAVE_WARC: bool = True
USE_WGET: bool = Field(default=lambda c: c.SAVE_WGET or c.SAVE_WARC)
WGET_BINARY: str = Field(default='wget')
WGET_ARGS: List[str] = [
'--no-verbose',
'--adjust-extension',
'--convert-links',
'--force-directories',
'--backup-converted',
'--span-hosts',
'--no-parent',
'-e', 'robots=off',
]
WGET_EXTRA_ARGS: List[str] = []
WGET_AUTO_COMPRESSION: bool = Field(default=True)
SAVE_WGET_REQUISITES: bool = Field(default=True)
WGET_USER_AGENT: str = Field(default='', alias='USER_AGENT')
WGET_TIMEOUT: int = Field(default=60, alias='TIMEOUT')
WGET_CHECK_SSL_VALIDITY: bool = Field(default=True, alias='CHECK_SSL_VALIDITY')
WGET_RESTRICT_FILE_NAMES: str = Field(default='windows', alias='RESTRICT_FILE_NAMES')
WGET_COOKIES_FILE: Optional[Path] = Field(default=None, alias='COOKIES_FILE')
@model_validator(mode='after')
def validate_use_ytdlp(self):
if self.USE_WGET and self.WGET_TIMEOUT < 10:
print(f'[red][!] Warning: TIMEOUT is set too low! (currently set to TIMEOUT={self.WGET_TIMEOUT} seconds)[/red]', file=sys.stderr)
print(' wget will fail to archive any sites if set to less than ~20 seconds.', file=sys.stderr)
print(' (Setting it somewhere over 60 seconds is recommended)', file=sys.stderr)
print(file=sys.stderr)
print(' If you want to disable media archiving entirely, set SAVE_MEDIA=False instead:', file=sys.stderr)
print(' https://github.com/ArchiveBox/ArchiveBox/wiki/Configuration#save_media', file=sys.stderr)
print(file=sys.stderr)
return self
# class WgetDependencyConfig(ConfigSet):
# WGET_BINARY: str = Field(default='wget')
# WGET_ARGS: Optional[List[str]] = Field(default=None)
# WGET_EXTRA_ARGS: List[str] = []
# WGET_DEFAULT_ARGS: List[str] = ['--timeout={TIMEOUT-10}']
# class WgetOptionsConfig(ConfigSet):
# # loaded from shared config
# WGET_AUTO_COMPRESSION: bool = Field(default=True)
# SAVE_WGET_REQUISITES: bool = Field(default=True)
# WGET_USER_AGENT: str = Field(default='', alias='USER_AGENT')
# WGET_TIMEOUT: int = Field(default=60, alias='TIMEOUT')
# WGET_CHECK_SSL_VALIDITY: bool = Field(default=True, alias='CHECK_SSL_VALIDITY')
# WGET_RESTRICT_FILE_NAMES: str = Field(default='windows', alias='RESTRICT_FILE_NAMES')
# WGET_COOKIES_FILE: Optional[Path] = Field(default=None, alias='COOKIES_FILE')
WGET_CONFIG = WgetConfig()
# CONFIG = {
# 'CHECK_SSL_VALIDITY': False,
# 'SAVE_WARC': False,
# 'TIMEOUT': 999,
# }
class WgetBinary(BaseBinary):
name: BinName = WGET_CONFIG.WGET_BINARY
binproviders_supported: List[InstanceOf[BinProvider]] = [apt, brew, env]
WGET_BINARY = WgetBinary()
# WGET_CONFIG = [
# WgetToggleConfig(**CONFIG),
# WgetDependencyConfig(**CONFIG),
# WgetOptionsConfig(**CONFIG),
# ]
class WgetExtractor(BaseExtractor):
name: ExtractorName = 'wget'
binary: str = WGET_BINARY.name
def get_output_path(self, snapshot) -> Path | None:
wget_index_path = wget_output_path(snapshot.as_link())
if wget_index_path:
return Path(wget_index_path)
return None
WGET_EXTRACTOR = WgetExtractor()
class WarcExtractor(BaseExtractor):
name: ExtractorName = 'warc'
binary: str = WGET_BINARY.name
# class WgetExtractor(Extractor):
# name: ExtractorName = 'wget'
# binary: Binary = WgetBinary()
# def get_output_path(self, snapshot) -> Path:
# return get_wget_output_path(snapshot)
# class WarcExtractor(Extractor):
# name: ExtractorName = 'warc'
# binary: Binary = WgetBinary()
# def get_output_path(self, snapshot) -> Path:
# return get_wget_output_path(snapshot)
def get_output_path(self, snapshot) -> Path | None:
warc_files = (snapshot.link_dir / 'warc').glob('*.warc.gz')
if warc_files:
return sorted(warc_files, key=lambda x: x.stat().st_size, reverse=True)[0]
return None
WARC_EXTRACTOR = WarcExtractor()
class WgetPlugin(BasePlugin):
app_label: str = 'wget'
verbose_name: str = 'WGET'
hooks: List[InstanceOf[BaseHook]] = []
hooks: List[InstanceOf[BaseHook]] = [
WGET_CONFIG,
WGET_BINARY,
WGET_EXTRACTOR,
WARC_EXTRACTOR,
]
PLUGIN = WgetPlugin()

View file

@ -1,10 +1,11 @@
import sys
from typing import List, Dict, ClassVar
from typing import List, Dict
from subprocess import run, PIPE
from rich import print
from pydantic import InstanceOf, Field, model_validator, AliasChoices
from pydantic_pkgr import BinProvider, BinName, BinProviderName, ProviderLookupDict
from abx.archivebox.base_plugin import BasePlugin
from abx.archivebox.base_configset import BaseConfigSet
from abx.archivebox.base_binary import BaseBinary, env, apt, brew