annotate bin/warc.py @ 39:8661062a50b1

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