Mercurial > hg > cc > cirrus_work

comparison bin/warc.py @ 39:8661062a50b1

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
moved from home bin
author Henry S. Thompson <ht@inf.ed.ac.uk>
date Wed, 05 Jul 2023 15:09:57 +0100
parents
children 689a0e311cd2
comparison
equal deleted inserted replaced
38:e3c440666f1a 39:8661062a50b1
1 #!/usr/bin/env python3
2 '''Stream a warc format file, invoking a callback on each record.
3 Callback can be limited by WARC-Type, record part'''
4 import sys,os,io
5
6 if (debug:=(sys.argv[1]=='-d')):
7 sys.argv.pop(1)
8
9 def warc(callback,types=['response'],whole=False,parts=7):
10 types=[(t if isinstance(t,bytes) else bytes(t,'utf8')) for t in types]
11 nb=0
12 stream=open(sys.argv[1],'rb',0)
13 bufsize=128*1024*1024
14 buf=bytearray(128*1024*1024)
15 l=b'\r\n'
16 while not stream.closed:
17 bp=0
18 while l==b'\r\n':
19 l=stream.readline()
20 nb+=(ln:=len(l))
21 if ln==0:
22 break
23 if l!=b'WARC/1.0\r\n':
24 raise ValueError("Not a WARC file? At %s: %s[%s]"%(nb-len(l),
25 l.decode('latin-1'),len(l)))
26 wtype=None
27 length=None
28 state=1
29 tr=None # Was this record truncated?
30 while l!=b'\r\n':
31 # WARC header
32 if parts & 1:
33 buf[bp:(bp:=bp+ln)]=l
34 l=stream.readline()
35 nb+=(ln:=len(l))
36 if l.startswith(b"Content-Length: "):
37 length=wl=int(l[16:].rstrip())
38 elif l.startswith(b"WARC-Truncated: "):
39 tr=l[16:].rstrip()
40 tr="EMPTY" if tr=="" else tr
41 elif l.startswith(b'WARC-Type: '):
42 wtype = l[11:-2]
43 start_2=bp
44 if (wtype in types):
45 if whole:
46 buf[bp:(bp:=bp+ln)]=l
47 elif (parts & 1):
48 callback(wtype,buf[:start_2],1)
49 if parts==1:
50 start_2=0
51 else:
52 start_2=bp
53 else:
54 start_2=0
55 bv=memoryview(buf)[start_2:start_2+length]
56 ii=0
57 while True and not stream.closed:
58 if (i:=stream.readinto(bv))==0:
59 break
60 ii+=i
61 if ii>=length:
62 break
63 bv=memoryview(buf)[start_2+ii:start_2+length]
64 if ii!=length:
65 raise ValueError("Chunk read losing, from %s got %s expected %s"%(nb,ii,length))
66 nb+=length
67 if wtype in types:
68 if whole:
69 callback(wtype,buf[0:start_2+length],7)
70 continue
71 # Only output parts (1 = WARC header, 2 = HTTP header, 4 = body) that are wanted
72 bl=None # for HTTP Content-Length for the length of the body?
73 L_start=start_2
74 state=2
75 bv=memoryview(buf)[start_2:start_2+length]
76 with io.BytesIO(bv) as rec_text:
77 for L in rec_text:
78 if state==2:
79 # HTTP header
80 wl -= len(L)
81 if not (L==b"" or L.startswith(b"\r")):
82 # Non-empty, it's (a continuation of) a header
83 if bl is None and L.startswith(b"Content-Length: "):
84 bl=int(L[16:].rstrip())
85 else:
86 # Blank line, HTTP header is finished
87 if parts & 2:
88 callback(wtype,buf[start_2:start_2+L_start],2)
89 state=4
90 # The above is just for sanity, because we do _not_
91 # continue with the outer loop,
92 # since we can now block-output the entire rest of the
93 # input buffer.
94 if bl is not None:
95 if bl!=wl:
96 print("length mismatch: %s %s %s here: %s given: %s trunc: %s"%\
97 (length,offset,filename,wl,bl,tr),file=sys.stderr)
98 # HTTP body
99 balance=start_2+rec_text.tell()
100 #print(balance,bl,wl,ll,ll-balance,file=sys.stderr)
101 # Output whatever is left
102 if parts & 4:
103 callback(wtype,buf[balance:balance+wl],4)
104 state=1
105
106 L_start=rec_text.tell()
107 OUT=open(sys.stdout.fileno(),'wb')
108
109 import re
110 TUPAT=re.compile(b'^WARC-Target-URI: (.*?)\r',re.MULTILINE)
111 LMPAT=re.compile(b'^Last-Modified: (.*?)\r',re.MULTILINE)
112
113 def showmeLMH(wtype,buf,part):
114 global URI
115 if part==1:
116 if (m:=TUPAT.search(buf)):
117 URI=m[1]
118 else:
119 raise ValueError(b"No target URI in %s ??"%buf)
120 else:
121 m=LMPAT.search(buf)
122 OUT.write(URI)
123 if m:
124 OUT.write(b'\t')
125 OUT.write(m[1])
126 OUT.write(b'\n')
127
128 def showme(wtype,buf,part):
129 if debug:
130 breakpoint()
131 OUT.write(b"%d\n%b"%(part,buf))
132
133 warc(showmeLMH,[b'response'],parts=3)
134
135 #warc(showme,[b'response','warcinfo','request','metadata'],int(sys.argv[2]))
136
137 #warc(showme,[b'response'],parts=int(sys.argv[2]))
138 #warc(showme,[b'response'],whole=True)