annotate bin/warc.py @ 50:55943918794e

a bit better
author Henry S. Thompson <ht@inf.ed.ac.uk>
date Fri, 07 Jul 2023 13:39:23 +0100
parents 699ef141af10
children c0b4359dd26a
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]
50
55943918794e a bit better
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 49
diff changeset
20 fpos=0
55943918794e a bit better
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 49
diff changeset
21 bl=stream.readinto(hdrBuf)
55943918794e a bit better
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 49
diff changeset
22 while True:
39
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
23 bp=0
48
d0d2fd9830d6 starting on conversion to direct-querying of buffer
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 46
diff changeset
24 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
25 bp+=2
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):
49
699ef141af10 just barely working for 1, need to rethink buffering
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 48
diff changeset
27 raise ValueError("Not a WARC file? At %s: %s[%s]"%(fpos,
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 bob=bp # in case 1 or whole
d0d2fd9830d6 starting on conversion to direct-querying of buffer
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 46
diff changeset
30 bp+=10
39
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
31 wtype=None
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
32 length=None
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
33 state=1
50
55943918794e a bit better
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 49
diff changeset
34 done=False
39
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
35 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
36 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
37 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
38 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
39 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
40 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
41 tr=l[bp+16:eol-2]
39
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
42 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
43 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
44 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
45 bp=eol
50
55943918794e a bit better
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 49
diff changeset
46 start_2=bp=eol+2
49
699ef141af10 just barely working for 1, need to rethink buffering
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 48
diff changeset
47 # need to read more if bp+length>hdrMax
39
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
48 if (wtype in types):
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
49 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
50 pass # buf[bp:(bp:=bp+ln)]=l
39
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
51 elif (parts & 1):
50
55943918794e a bit better
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 49
diff changeset
52 print('cb')
55943918794e a bit better
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 49
diff changeset
53 OUT=callback(wtype,buf[bob:eol],1)
55943918794e a bit better
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 49
diff changeset
54 sys.stdout.flush()
55943918794e a bit better
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 49
diff changeset
55 if parts!=1:
55943918794e a bit better
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 49
diff changeset
56 # everything from bv= goes here
55943918794e a bit better
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 49
diff changeset
57 pass
55943918794e a bit better
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 49
diff changeset
58 print(wtype,fpos,bp,bp-bob,length)
55943918794e a bit better
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 49
diff changeset
59 stream.seek(fpos:=fpos+(bp-bob)+length)
55943918794e a bit better
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 49
diff changeset
60 print(fpos)
55943918794e a bit better
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 49
diff changeset
61 if done:
55943918794e a bit better
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 49
diff changeset
62 return
55943918794e a bit better
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 49
diff changeset
63 buf[0:hdrMax-fpos]=buf[fpos:hdrMax]
55943918794e a bit better
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 49
diff changeset
64 n=stream.readinto(memoryview(buf)[fpos:hdrMax])
55943918794e a bit better
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 49
diff changeset
65 if n<hdrMax-fpos or n==0:
55943918794e a bit better
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 49
diff changeset
66 done=True
55943918794e a bit better
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 49
diff changeset
67 #while not buf.startswith(b'\r\n',bp):
55943918794e a bit better
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 49
diff changeset
68 OUT.write(b"=====\n")
55943918794e a bit better
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 49
diff changeset
69 OUT.write(buf[0:100])
55943918794e a bit better
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 49
diff changeset
70 continue
55943918794e a bit better
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 49
diff changeset
71 return
39
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
72 bv=memoryview(buf)[start_2:start_2+length]
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
73 ii=0
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
74 while True and not stream.closed:
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
75 if (i:=stream.readinto(bv))==0:
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
76 break
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
77 ii+=i
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
78 if ii>=length:
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
79 break
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
80 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
81 if ii!=length:
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
82 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
83 nb+=length
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
84 if wtype in types:
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
85 if whole:
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
86 callback(wtype,buf[0:start_2+length],7)
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
87 continue
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
88 # 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
89 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
90 L_start=start_2
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
91 state=2
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
92 bv=memoryview(buf)[start_2:start_2+length]
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
93 with io.BytesIO(bv) as rec_text:
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
94 for L in rec_text:
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
95 if state==2:
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
96 # HTTP header
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
97 wl -= len(L)
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
98 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
99 # 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
100 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
101 bl=int(L[16:].rstrip())
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
102 else:
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
103 # Blank line, HTTP header is finished
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
104 if parts & 2:
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
105 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
106 state=4
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
107 # 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
108 # continue with the outer loop,
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
109 # 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
110 # input buffer.
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
111 if bl is not None:
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
112 if bl!=wl:
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
113 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
114 (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
115 # HTTP body
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
116 balance=start_2+rec_text.tell()
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
117 #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
118 # Output whatever is left
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
119 if parts & 4:
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
120 callback(wtype,buf[balance:balance+wl],4)
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
121 state=1
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
122
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
123 L_start=rec_text.tell()