annotate bin/warc.py @ 45:212da3fe3b19

make test 1 idempotent
author Henry S. Thompson <ht@inf.ed.ac.uk>
date Wed, 05 Jul 2023 19:32:02 +0100
parents 69be1131bcc5
children 44d3a4f4ea51
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
42
689a0e311cd2 make warc.py a library, separate out testing
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 39
diff changeset
2 '''Stream a gzipped warc format file, invoking a callback on each record.
39
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
42
689a0e311cd2 make warc.py a library, separate out testing
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 39
diff changeset
5 import sys,io
689a0e311cd2 make warc.py a library, separate out testing
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 39
diff changeset
6 from isal import igzip
39
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
7
42
689a0e311cd2 make warc.py a library, separate out testing
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 39
diff changeset
8 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
9 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
10 nb=0
42
689a0e311cd2 make warc.py a library, separate out testing
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 39
diff changeset
11 stream=open(filename,'rb',0)
39
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
12 bufsize=128*1024*1024
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
13 buf=bytearray(128*1024*1024)
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
14 l=b'\r\n'
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
15 while not stream.closed:
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
16 bp=0
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
17 while l==b'\r\n':
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
18 l=stream.readline()
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
19 nb+=(ln:=len(l))
43
69be1131bcc5 get EOF right, finally
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 42
diff changeset
20 if ln==0:
69be1131bcc5 get EOF right, finally
Henry S. Thompson <ht@inf.ed.ac.uk>
parents: 42
diff changeset
21 break
39
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
22 if l!=b'WARC/1.0\r\n':
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
23 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
24 l.decode('latin-1'),len(l)))
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
25 wtype=None
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
26 length=None
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
27 state=1
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
28 tr=None # Was this record truncated?
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
29 while l!=b'\r\n':
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
30 # WARC header
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
31 if parts & 1:
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
32 buf[bp:(bp:=bp+ln)]=l
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
33 l=stream.readline()
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
34 nb+=(ln:=len(l))
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
35 if l.startswith(b"Content-Length: "):
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
36 length=wl=int(l[16:].rstrip())
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
37 elif l.startswith(b"WARC-Truncated: "):
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
38 tr=l[16:].rstrip()
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
39 tr="EMPTY" if tr=="" else tr
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
40 elif l.startswith(b'WARC-Type: '):
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
41 wtype = l[11:-2]
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
42 start_2=bp
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
43 if (wtype in types):
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
44 if whole:
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
45 buf[bp:(bp:=bp+ln)]=l
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
46 elif (parts & 1):
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
47 callback(wtype,buf[:start_2],1)
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
48 if parts==1:
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
49 start_2=0
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
50 else:
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
51 start_2=bp
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
52 else:
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
53 start_2=0
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
54 bv=memoryview(buf)[start_2:start_2+length]
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
55 ii=0
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
56 while True and not stream.closed:
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
57 if (i:=stream.readinto(bv))==0:
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
58 break
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
59 ii+=i
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
60 if ii>=length:
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
61 break
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
62 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
63 if ii!=length:
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
64 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
65 nb+=length
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
66 if wtype in types:
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
67 if whole:
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
68 callback(wtype,buf[0:start_2+length],7)
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
69 continue
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
70 # 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
71 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
72 L_start=start_2
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
73 state=2
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
74 bv=memoryview(buf)[start_2:start_2+length]
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
75 with io.BytesIO(bv) as rec_text:
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
76 for L in rec_text:
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
77 if state==2:
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
78 # HTTP header
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
79 wl -= len(L)
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
80 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
81 # 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
82 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
83 bl=int(L[16:].rstrip())
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
84 else:
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
85 # Blank line, HTTP header is finished
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
86 if parts & 2:
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
87 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
88 state=4
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
89 # 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
90 # continue with the outer loop,
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
91 # 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
92 # input buffer.
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
93 if bl is not None:
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
94 if bl!=wl:
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
95 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
96 (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
97 # HTTP body
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
98 balance=start_2+rec_text.tell()
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
99 #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
100 # Output whatever is left
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
101 if parts & 4:
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
102 callback(wtype,buf[balance:balance+wl],4)
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
103 state=1
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
104
8661062a50b1 moved from home bin
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
105 L_start=rec_text.tell()