Mercurial > hg > cc > cirrus_work

annotate bin/ix.py @ 18:046dbe557911

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
write to tmp file implemented
author Henry S. Thompson <ht@inf.ed.ac.uk>
date Sun, 07 Aug 2022 13:58:33 +0100
parents
children cbac7dfe2f24
Ignore whitespace changes - Everywhere: Within whitespace: At end of lines:
rev   line source
18
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
1 #!/usr/bin/env python3
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
2 '''Extract request records from Common Crawl WARC-format files
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
3 given length, offset and filename triples.
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
4 Input one triple on command line, or
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
5 triples from stdin as tab-delimited lines
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
6 or complete cdx index lines.
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
7 In all cases by 'filename' is meant crawlid/segmentid/type/filename
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
8
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
9 Note that if no output flag(s) is/are given, the whole WARC record will be output, more efficiently than would be the case if -whb is given.'''
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
10
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
11 import sys, argparse, regex, os, shutil, io, gzip, time, shlex
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
12 from isal import igzip
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
13 from subprocess import Popen, PIPE
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
14 #import asyncio
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
15
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
16 HACK_USAGE=regex.compile('\[-x\]\n\s*\[length\] \[offset\] \[filename\]')
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
17 BINOUT=sys.stdout.buffer
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
18 FPAT="/%s/%s/orig/%s/%s"
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
19
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
20 CMD_PROC=None
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
21 TMPFILENAME=None
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
22
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
23 class HackFormat(argparse.RawDescriptionHelpFormatter):
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
24 def format_help(self):
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
25 FOO=argparse.RawDescriptionHelpFormatter.format_help(self)
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
26 return HACK_USAGE.sub('\n [ ( -x | length offset filename ) ]',
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
27 FOO)
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
28
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
29 def process(options,buf,filename,offset,length,whole):
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
30 global CMD_PROC, BINOUT, TMPFILENAME, TMPFILE
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
31 if options.save:
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
32 (tf,TMPFILENAME)=tempfile.mkstemp()
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
33 TMPFILE=open(tf,mode='wb')
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
34 if options.cmd:
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
35 CMD_PROC=Popen(shlex.split(options.cmd),stdin=PIPE,bufsize=0)
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
36 BINOUT=CMD_PROC.stdin
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
37 process1(options,buf,filename,offset,length,whole)
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
38 if options.save:
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
39 TMPFILE.close()
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
40 if options.cmd:
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
41 _output_subproc(bytes(TMPFILENAME,'utf-8'))
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
42 _output_subproc(b"\n")
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
43 else:
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
44 BINOUT.write(bytes(TMPFILENAME,'utf-8'))
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
45 BINOUT.write(b"\n")
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
46 if options.cmd:
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
47 # Wind up subproc
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
48 BINOUT.close()
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
49 if CMD_PROC.wait()!=0: # could/should be async?
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
50 print("subproc of %s:%s:%s failed with %s"%(length,offset,filename,
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
51 CMD_PROC.returncode),
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
52 file=sys.stderr)
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
53 if options.save:
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
54 # not if async?
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
55 os.unlink(TMPFILENAME)
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
56 TMPFILENAME=None
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
57 elif options.save:
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
58 print("%s will need to be deleted"%TMPFILENAME,file=sys.stderr)
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
59 TMPFILENAME=None
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
60
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
61 def _output_tmpfile(buf):
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
62 TMPFILE.write(buf)
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
63
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
64 def _output_stdout(buf):
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
65 BINOUT.write(buf)
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
66
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
67 def _output_subproc(buf):
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
68 toWrite=len(buf)
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
69 while toWrite>0:
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
70 toWrite -= BINOUT.write(buf)
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
71
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
72 def process1(options,buf,filename,offset,length,whole):
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
73 root=options.root
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
74 rfn=root+filename
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
75 if root!="/beegfs/common_crawl":
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
76 # Support using ramdisk or other local disk as a faster cached
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
77 if not os.path.exists(rfn):
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
78 if not os.path.exists(os.path.dirname(rfn)):
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
79 os.makedirs(os.path.dirname(rfn))
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
80 with io.FileIO('/beegfs/common_crawl'+filename,'r') as infile, \
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
81 io.FileIO(rfn,'w') as outfile:
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
82 #shutil.copyfileobj(infile,outfile,128*1024*1024)
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
83 while True:
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
84 l=infile.readinto(buf)
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
85 if l==0:
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
86 break
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
87 outfile.write(memoryview(buf)[:l])
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
88 file=open(rfn,'rb',0)
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
89 file.seek(offset)
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
90 bv=memoryview(buf)[:length]
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
91 nb=file.readinto(bv)
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
92 file.close()
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
93 if nb!=length:
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
94 raise ValueError("Chunk read losing: %s, got %s expected %s at %s"%(file.name,
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
95 nb,length,offset))
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
96 if whole and options.zipped:
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
97 _output(bv)
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
98 return
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
99 gzip_chunk = io.BytesIO(bv)
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
100 uv=memoryview(buf)[length:]
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
101 with igzip.IGzipFile(fileobj=gzip_chunk) as gzip_fin:
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
102 ll=0
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
103 while True:
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
104 l=gzip_fin.readinto(uv)
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
105 if not l:
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
106 break
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
107 ll+=l
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
108 cb=memoryview(uv)[:ll]
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
109 if whole:
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
110 _output(cb)
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
111 return
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
112 # Only output parts (0 = WARC header, 1 = HTTP header, 2 = body) that are wanted
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
113 state=0
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
114 tr=None # Was this record truncated?
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
115 bl=None # for HTTP Content-Length for the length of the body?
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
116 with io.BytesIO(cb) as clear_text:
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
117 for L in clear_text:
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
118 if state==0:
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
119 # WARC header
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
120 if L.startswith(b"Content-Length: "):
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
121 wl=int(L[16:].rstrip())
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
122 elif L.startswith(b"WARC-Truncated: "):
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
123 tr=L[16:].rstrip()
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
124 tr="EMPTY" if tr=="" else tr
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
125 elif L==b"" or L.startswith(b"\r"): # for idempotency
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
126 # Blank line, WARC header is finished
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
127 if not (options.headers or options.body):
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
128 return
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
129 state=1
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
130 # Note we preserve the empty line
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
131 if options.warc:
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
132 _output(L)
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
133 continue
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
134 if state==1:
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
135 # HTTP header
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
136 wl -= len(L)
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
137 if not (L==b"" or L.startswith(b"\r")):
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
138 # Non-blank, it's a header
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
139 if bl is None and L.startswith(b"Content-Length: "):
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
140 bl=int(L[16:].rstrip())
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
141 if options.headers:
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
142 _output(L)
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
143 else:
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
144 # Blank line, HTTP header is finished
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
145 if not options.body:
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
146 return
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
147 if options.headers:
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
148 _output(L)
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
149 state=2
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
150 # The above is just for sanity, because we do _not_
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
151 # continue with the outer loop,
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
152 # since we can now block-output the entire rest of the
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
153 # input buffer.
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
154 if bl is not None:
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
155 if bl!=wl:
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
156 print("length mismatch: %s %s %s here: %s given: %s trunc: %s"%\
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
157 (length,offset,filename,wl,bl,tr),file=sys.stderr)
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
158 # HTTP body
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
159 balance=clear_text.tell()
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
160 #print(balance,bl,wl,ll,ll-balance,file=sys.stderr)
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
161 # Output whatever is left
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
162 _output(cb[balance:balance+wl])
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
163 return
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
164
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
165 def main():
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
166 global _output,TMPFILE,TMPFILENAME,tempfile
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
167 parser = argparse.ArgumentParser(
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
168 description='''Extract records from warc files given length, offset and file triples.
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
169 Input one triple on command line, or
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
170 triples from stdin as tab-delimited lines
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
171 or complete cdx index lines.
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
172 In all cases by 'filename' is meant crawlid/segmentid/type/filename''',
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
173 epilog='''Note that if no output flag(s) is/are given,
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
174 the whole WARC record will be output, more efficiently than
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
175 would be the case if all three flags were given.''',
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
176 add_help=False,
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
177 conflict_handler='resolve',
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
178 formatter_class=HackFormat
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
179 )
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
180 fphelp=('format string for turning 4 filename components into a path, must contain %%s exactly 4 times,\ndefault is "%s"'%FPAT).replace('%s','%%s')
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
181 parser.add_argument('--help',help='Show help',action='help')
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
182 parser.add_argument('-d','--debug',help='Debug output',action='store_true')
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
183 parser.add_argument('-w','--warc',help='output WARC headers',
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
184 action='store_true')
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
185 parser.add_argument('-h','--headers',help='output HTTP headers',
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
186 action='store_true')
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
187 parser.add_argument('-b','--body',help='output HTTP body',
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
188 action='store_true')
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
189 parser.add_argument('-c','--cmd',help='pipes each result thru CMD')
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
190 parser.add_argument('-m','--module.function',help='module.function to call with a stream'),
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
191 parser.add_argument('-s','--save',action='store_true',
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
192 help="write to a temporary file and output the name")
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
193 parser.add_argument('-f','--fpath',
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
194 help=fphelp,
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
195 default=FPAT)
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
196 parser.add_argument('-r','--root',nargs='?',
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
197 help='File path root, create a copy there if necessary',
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
198 default='/beegfs/common_crawl'),
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
199 parser.add_argument('-z','--zipped',
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
200 help="output raw gzipped record, ignored if any of -bhw supplied",
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
201 action='store_true')
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
202 sg=parser.add_mutually_exclusive_group()
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
203 sg.add_argument('-x','--index',
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
204 help='take lines of triples from a cdx index file as input',
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
205 action='store_true')
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
206 sg.add_argument('length',type=int,
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
207 help='length in bytes of gzipped record',
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
208 nargs='?')
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
209 parser.add_argument('offset',type=int,
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
210 help='start position in bytes of gzipped record',
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
211 nargs='?')
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
212 parser.add_argument('filename',
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
213 help='pathname of gzipped Common Crawl WARC-format file',
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
214 nargs='?')
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
215 # Hack the order of optional and positional in the help output
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
216 parser._action_groups.sort(key=lambda g:g.title)
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
217 #parser.print_help()
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
218 pa=parser.parse_args(sys.argv[1:])
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
219 #print(pa,file=sys.stderr)
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
220 if pa.length is not None:
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
221 # We have to enforce our own check..
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
222 if pa.offset is None or pa.filename is None:
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
223 parser.error("length, offset and filename must all be supplied together")
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
224
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
225 buf=bytearray(128*1024*1024)
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
226
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
227 whole=not (pa.warc or pa.headers or pa.body)
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
228 if pa.save:
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
229 _output=_output_tmpfile
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
230 import tempfile
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
231 elif pa.cmd:
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
232 _output = _output_subproc
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
233 else:
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
234 _output = _output_stdout
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
235 if pa.index:
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
236 CDX=regex.compile('length": "([0-9]*)", "offset": "([0-9]*)", "filename": "crawl-data/([^/]*)/segments/([^/]*)/(warc|crawldiagnostics)/(.*\.gz)"') # no robotstxt yet...
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
237 for l in sys.stdin:
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
238 m=CDX.search(l)
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
239 if m is None:
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
240 if l.find('/robotstxt/')>-1:
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
241 continue
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
242 print("index line problem: \"%s\""%l,file=sys.stderr,end='')
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
243 exit(2)
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
244 f=pa.fpath%(m[3:7])
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
245 try:
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
246 process(pa,buf,f,
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
247 int(m[2]),int(m[1]),whole)
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
248 except Exception as e:
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
249 if pa.debug:
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
250 import traceback
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
251 traceback.print_exc(file=sys.stderr)
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
252 else:
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
253 print("Process fail: %s, input line:\n %s"%(e,l),
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
254 file=sys.stderr,end='')
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
255 exit(3)
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
256 elif pa.length is not None:
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
257 print(pa.filename,file=sys.stderr)
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
258 process(pa,buf,pa.fpath%tuple(pa.filename.split('/')),
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
259 pa.offset,pa.length,whole)
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
260 else:
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
261 print("Reading length, offset, filename tab-delimited triples from stdin...",
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
262 file=sys.stderr)
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
263 for l in sys.stdin:
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
264 try:
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
265 (length,offset,filename)=l.rstrip().split('\t')
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
266 length=int(length)
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
267 offset=int(offset)
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
268 except ValueError as e:
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
269 parser.error('Invalid input line: %s\n "%s"'%(e,l))
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
270 process(pa,buf,pa.fpath%tuple(filename.split('/')),
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
271 offset,length,whole)
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
272
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
273 if __name__ == "__main__":
046dbe557911 write to tmp file implemented
Henry S. Thompson <ht@inf.ed.ac.uk>
parents:
diff changeset
274 main()