annotate bin/warc.py @ 53:0dc144bd027c

made 1 mean 1, still losing after a while
author Henry S. Thompson <ht@inf.ed.ac.uk>
date Fri, 07 Jul 2023 19:04:16 +0100
parents c0b4359dd26a
children 9c63039a9b6d
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
46
44d3a4f4ea51 support on-board unzipping, reduce buffer size to 2MB
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 43
diff changeset
2 '''Stream a warc format file, unzipping if necessary, invoking a
44d3a4f4ea51 support on-board unzipping, reduce buffer size to 2MB
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 43
diff changeset
3 callback on each record. Callback can be limited by WARC-Type, record
44d3a4f4ea51 support on-board unzipping, reduce buffer size to 2MB
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 43
diff changeset
4 part'''
39
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
5
42
689a0e311cd2 make warc.py a library, separate out testing
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 39
diff changeset
6 import sys,io
689a0e311cd2 make warc.py a library, separate out testing
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 39
diff changeset
7 from isal import igzip
39
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
8
42
689a0e311cd2 make warc.py a library, separate out testing
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 39
diff changeset
9 def warc(filename,callback,types=['response'],whole=False,parts=7,debug=False):
39
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
46
44d3a4f4ea51 support on-board unzipping, reduce buffer size to 2MB
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 43
diff changeset
12 if filename.endswith(".gz"):
44d3a4f4ea51 support on-board unzipping, reduce buffer size to 2MB
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 43
diff changeset
13 stream=igzip.IGzipFile(filename=filename)
44d3a4f4ea51 support on-board unzipping, reduce buffer size to 2MB
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 43
diff changeset
14 else:
44d3a4f4ea51 support on-board unzipping, reduce buffer size to 2MB
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 43
diff changeset
15 stream=open(filename,'rb',0)
48
d0d2fd9830d6 starting on conversion to direct-querying of buffer
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 46
diff changeset
16 bufSize=2*1024*1024
d0d2fd9830d6 starting on conversion to direct-querying of buffer
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 46
diff changeset
17 hdrMax=16*1024
d0d2fd9830d6 starting on conversion to direct-querying of buffer
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 46
diff changeset
18 buf=bytearray(bufSize)
d0d2fd9830d6 starting on conversion to direct-querying of buffer
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 46
diff changeset
19 hdrBuf=memoryview(buf)[:hdrMax]
51
c0b4359dd26a working better, gets confused by 3-part response
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 50
diff changeset
20 fpos=bl=stream.readinto(hdrBuf)
50
55943918794e a bit better
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 49
diff changeset
21 while True:
39
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
22 bp=0
48
d0d2fd9830d6 starting on conversion to direct-querying of buffer
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 46
diff changeset
23 while buf.startswith(b'\r\n',bp):
d0d2fd9830d6 starting on conversion to direct-querying of buffer
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 46
diff changeset
24 bp+=2
53
0dc144bd027c made 1 mean 1, still losing after a while
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 51
diff changeset
25 bob=bp
48
d0d2fd9830d6 starting on conversion to direct-querying of buffer
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 46
diff changeset
26 if not buf.startswith(b'WARC/1.0\r\n',bp):
51
c0b4359dd26a working better, gets confused by 3-part response
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 50
diff changeset
27 raise ValueError("Not a WARC file? At %s: %s[%s]"%(bp,
48
d0d2fd9830d6 starting on conversion to direct-querying of buffer
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 46
diff changeset
28 buf[bp:min(bl,bp+20)].decode('latin-1'), bl-bp))
d0d2fd9830d6 starting on conversion to direct-querying of buffer
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 46
diff changeset
29 bp+=10
39
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
30 wtype=None
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
31 length=None
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
32 state=1
50
55943918794e a bit better
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 49
diff changeset
33 done=False
39
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
34 tr=None # Was this record truncated?
48
d0d2fd9830d6 starting on conversion to direct-querying of buffer
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 46
diff changeset
35 while not buf.startswith(b'\r\n',bp):
d0d2fd9830d6 starting on conversion to direct-querying of buffer
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 46
diff changeset
36 eol=buf.index(b'\r\n',bp)+2
d0d2fd9830d6 starting on conversion to direct-querying of buffer
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 46
diff changeset
37 if buf.startswith(b"Content-Length: ",bp):
d0d2fd9830d6 starting on conversion to direct-querying of buffer
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 46
diff changeset
38 length=wl=int(buf[bp+16:eol-2])
d0d2fd9830d6 starting on conversion to direct-querying of buffer
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 46
diff changeset
39 elif buf.startswith(b"WARC-Truncated: ",bp):
d0d2fd9830d6 starting on conversion to direct-querying of buffer
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 46
diff changeset
40 tr=l[bp+16:eol-2]
39
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
41 tr="EMPTY" if tr=="" else tr
48
d0d2fd9830d6 starting on conversion to direct-querying of buffer
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 46
diff changeset
42 elif buf.startswith(b'WARC-Type: ',bp):
d0d2fd9830d6 starting on conversion to direct-querying of buffer
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 46
diff changeset
43 wtype = bytes(buf[bp+11:eol-2])
d0d2fd9830d6 starting on conversion to direct-querying of buffer
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 46
diff changeset
44 bp=eol
51
c0b4359dd26a working better, gets confused by 3-part response
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 50
diff changeset
45 bp=eol+2
39
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
46 if (wtype in types):
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
47 if whole:
49
699ef141af10 just barely working for 1, need to rethink buffering
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 48
diff changeset
48 pass # buf[bp:(bp:=bp+ln)]=l
39
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
49 elif (parts & 1):
50
55943918794e a bit better
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 49
diff changeset
50 print('cb')
53
0dc144bd027c made 1 mean 1, still losing after a while
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 51
diff changeset
51 OUT=callback(wtype,buf[bob:eol],1)
50
55943918794e a bit better
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 49
diff changeset
52 if parts!=1:
55943918794e a bit better
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 49
diff changeset
53 # everything from bv= goes here
51
c0b4359dd26a working better, gets confused by 3-part response
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 50
diff changeset
54 # need to read more if eol+length>hdrMax
50
55943918794e a bit better
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 49
diff changeset
55 pass
51
c0b4359dd26a working better, gets confused by 3-part response
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 50
diff changeset
56 print(wtype,bob,eol,length)
c0b4359dd26a working better, gets confused by 3-part response
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 50
diff changeset
57 jumpTo=bp+length
c0b4359dd26a working better, gets confused by 3-part response
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 50
diff changeset
58 buf[0:hdrMax-jumpTo]=buf[jumpTo:hdrMax]
c0b4359dd26a working better, gets confused by 3-part response
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 50
diff changeset
59 stream.seek(fpos:=fpos+jumpTo)
50
55943918794e a bit better
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 49
diff changeset
60 if done:
51
c0b4359dd26a working better, gets confused by 3-part response
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 50
diff changeset
61 continue
c0b4359dd26a working better, gets confused by 3-part response
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 50
diff changeset
62 n=stream.readinto(memoryview(buf)[hdrMax-jumpTo:hdrMax])
c0b4359dd26a working better, gets confused by 3-part response
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 50
diff changeset
63 print('read',n)
c0b4359dd26a working better, gets confused by 3-part response
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 50
diff changeset
64 if n<jumpTo or n==0:
c0b4359dd26a working better, gets confused by 3-part response
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 50
diff changeset
65 print('done',n,jumpTo)
50
55943918794e a bit better
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 49
diff changeset
66 done=True
51
c0b4359dd26a working better, gets confused by 3-part response
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 50
diff changeset
67 bp=0
50
55943918794e a bit better
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 49
diff changeset
68 #while not buf.startswith(b'\r\n',bp):
55943918794e a bit better
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 49
diff changeset
69 OUT.write(b"=====\n")
55943918794e a bit better
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 49
diff changeset
70 OUT.write(buf[0:100])
55943918794e a bit better
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 49
diff changeset
71 continue
55943918794e a bit better
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 49
diff changeset
72 return
39
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
73 bv=memoryview(buf)[start_2:start_2+length]
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
74 ii=0
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
75 while True and not stream.closed:
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
76 if (i:=stream.readinto(bv))==0:
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
77 break
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
78 ii+=i
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
79 if ii>=length:
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
80 break
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
81 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
82 if ii!=length:
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
83 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
84 nb+=length
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
85 if wtype in types:
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
86 if whole:
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
87 callback(wtype,buf[0:start_2+length],7)
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
88 continue
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
89 # 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
90 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
91 L_start=start_2
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
92 state=2
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
93 bv=memoryview(buf)[start_2:start_2+length]
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
94 with io.BytesIO(bv) as rec_text:
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
95 for L in rec_text:
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
96 if state==2:
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
97 # HTTP header
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
98 wl -= len(L)
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
99 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
100 # 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
101 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
102 bl=int(L[16:].rstrip())
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
103 else:
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
104 # Blank line, HTTP header is finished
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
105 if parts & 2:
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
106 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
107 state=4
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
108 # 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
109 # continue with the outer loop,
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
110 # 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
111 # input buffer.
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
112 if bl is not None:
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
113 if bl!=wl:
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
114 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
115 (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
116 # HTTP body
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
117 balance=start_2+rec_text.tell()
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
118 #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
119 # Output whatever is left
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
120 if parts & 4:
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
121 callback(wtype,buf[balance:balance+wl],4)
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
122 state=1
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
123
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
124 L_start=rec_text.tell()