headers.py 2.0 KB

1234567891011121314151617181920212223242526272829303132333435363738394041424344454647484950515253545556575859606162636465666768697071727374757677787980
  1. __package__ = 'archivebox.extractors'
  2. from pathlib import Path
  3. from typing import Optional
  4. from django.db.models import Model
  5. from ..index.schema import ArchiveResult, ArchiveOutput
  6. from ..system import atomic_write
  7. from ..util import (
  8. enforce_types,
  9. get_headers,
  10. )
  11. from ..config import (
  12. TIMEOUT,
  13. CURL_BINARY,
  14. CURL_ARGS,
  15. CURL_USER_AGENT,
  16. CURL_VERSION,
  17. CHECK_SSL_VALIDITY,
  18. SAVE_HEADERS
  19. )
  20. from ..logging_util import TimedProgress
  21. # output = 'headers.json'
  22. @enforce_types
  23. def should_save_headers(snapshot: Model, overwrite: Optional[bool]=False, out_dir: Optional[str]=None) -> bool:
  24. out_dir = out_dir or snapshot.snapshot_dir
  25. if not SAVE_HEADERS:
  26. return False
  27. if overwrite:
  28. return True
  29. output = Path(out_dir or snapshot.snapshot_dir) / 'headers.json'
  30. return not output.exists()
  31. @enforce_types
  32. def save_headers(snapshot: Model, out_dir: Optional[str]=None, timeout: int=TIMEOUT) -> ArchiveResult:
  33. """Download site headers"""
  34. out_dir = Path(out_dir or snapshot.snapshot_dir)
  35. output_folder = out_dir.absolute()
  36. output: ArchiveOutput = 'headers.json'
  37. status = 'succeeded'
  38. timer = TimedProgress(timeout, prefix=' ')
  39. cmd = [
  40. CURL_BINARY,
  41. *CURL_ARGS,
  42. '--head',
  43. '--max-time', str(timeout),
  44. *(['--user-agent', '{}'.format(CURL_USER_AGENT)] if CURL_USER_AGENT else []),
  45. *([] if CHECK_SSL_VALIDITY else ['--insecure']),
  46. snapshot.url,
  47. ]
  48. try:
  49. json_headers = get_headers(snapshot.url, timeout=timeout)
  50. output_folder.mkdir(exist_ok=True)
  51. atomic_write(str(output_folder / "headers.json"), json_headers)
  52. except (Exception, OSError) as err:
  53. status = 'failed'
  54. output = err
  55. finally:
  56. timer.end()
  57. return ArchiveResult(
  58. cmd=cmd,
  59. pwd=str(out_dir),
  60. cmd_version=CURL_VERSION,
  61. output=output,
  62. status=status,
  63. **timer.stats,
  64. )