changeset 175:d123ef7fdb82

working on implementing types and parts: 1, 2, 4 working, 3 not
author Henry S. Thompson <ht@inf.ed.ac.uk>
date Mon, 03 Jul 2023 18:16:14 +0100
parents bfe9085a1d39
children 97137f5bbe0f
files bin/warc.py
diffstat 1 files changed, 84 insertions(+), 88 deletions(-) [+]
line wrap: on
line diff
--- a/bin/warc.py	Tue Jan 10 17:48:26 2023 +0000
+++ b/bin/warc.py	Mon Jul 03 18:16:14 2023 +0100
@@ -1,126 +1,122 @@
 #!/usr/bin/env python3
-'''Stream a warc format file, invoking a callback on each part.
-Callback can be limited by WARC-Type'''
-import sys,os
+'''Stream a warc format file, invoking a callback on each record.
+Callback can be limited by WARC-Type, record part'''
+import sys,os,io
 
-def warc(callback,types=['response']):
+if (debug:=(sys.argv[1]=='-d')):
+  sys.argv.pop(1)
+
+def warc(callback,types=['response'],parts=7):
+  types=[(t if isinstance(t,bytes) else bytes(t,'utf8')) for t in types]
   nb=0
   stream=open(sys.argv[1],'rb',0)
   bufsize=128*1024*1024
   buf=bytearray(128*1024*1024)
   l=b'\r\n'
   while True:
+    bp=0
     while l==b'\r\n':
       l=stream.readline()
-      nb+=len(l)
+      nb+=(ln:=len(l))
     if l!=b'WARC/1.0\r\n':
-      if l==b'':
-        return
       raise ValueError("Not a WARC file? At %s: %s[%s]"%(nb-len(l),
                                                          l.decode('latin-1'),len(l)))
     wtype=None
     length=None
+    state=1
+    tr=None # Was this record truncated?
     while l!=b'\r\n':
+      if parts & 1:
+        buf[bp:(bp:=bp+ln)]=l
       l=stream.readline()
-      nb+=len(l)
-      if l.startswith(b'WARC-Type: '):
+      nb+=(ln:=len(l))
+      # WARC header
+      if l.startswith(b"Content-Length: "):
+        length=wl=int(l[16:].rstrip())
+      elif l.startswith(b"WARC-Truncated: "):
+        tr=l[16:].rstrip()
+        tr="EMPTY" if tr=="" else tr
+      elif l.startswith(b'WARC-Type: '):
         wtype = l[11:-2]
-      elif l.startswith(b'Content-Length: '):
-        length = int(l[16:])
-    bv=memoryview(buf)[:length]
+    start_2=bp
+    if (wtype in types) and (parts & 1):
+      if parts!=1:
+        buf[bp:(bp:=bp+ln)]=l
+        start_2=bp
+      if parts!=7:
+        callback(wtype,buf[:start_2],1)
+    else:
+      start_2=0
+    bv=memoryview(buf)[start_2:start_2+length]
     ii=0
-    while True:
-      i=stream.readinto(bv)
+    while True and not stream.closed:
+      if (i:=stream.readinto(bv))==0:
+        break
       ii+=i
       if ii>=length:
         break
-      bv=memoryview(buf)[ii:length]
+      bv=memoryview(buf)[start_2+ii:start_2+length]
     if ii!=length:
       raise ValueError("Chunk read losing, from %s got %s expected %s"%(nb,ii,length))
     nb+=length
+    bv=memoryview(buf)[start_2:start_2+length]
     if wtype in types:
-      callback(wtype,memoryview(buf[:length]))
-  if whole and options.zipped:
-    _output(bv)
-    return
-  gzip_chunk = io.BytesIO(bv)
-  uv=memoryview(buf)[length:]
-  with igzip.IGzipFile(fileobj=gzip_chunk) as gzip_fin:
-    ll=0
-    while True:
-      l=gzip_fin.readinto(uv)
-      if not l:
-        break
-      ll+=l
-    cb=memoryview(uv)[:ll]
-    if whole:
-      _output(cb)
-      return
-  # Only output parts (0 = WARC header, 1 = HTTP header, 2 = body) that are wanted
-  state=0
-  tr=None # Was this record truncated?
-  bl=None # for HTTP Content-Length for the length of the body?
-  with io.BytesIO(cb) as clear_text:
-    for L in clear_text:
-      if state==0:
-        # WARC header
-        if L.startswith(b"Content-Length: "):
-          wl=int(L[16:].rstrip())
-        elif L.startswith(b"WARC-Truncated: "):
-          tr=L[16:].rstrip()
-          tr="EMPTY" if tr=="" else tr
-        elif L==b"" or L.startswith(b"\r"): # for idempotency
-          # Blank line, WARC header is finished
-          if not (options.headers or options.body):
-            return
-          state=1
-          # Note we preserve the empty line
-        if options.warc:
-          _output(L)
+      if parts==7:
+        callback(wtype,memoryview(buf)[0:start_2+length],7)
         continue
-      if state==1:
-        # HTTP header
-        wl -= len(L)
-        if not (L==b"" or L.startswith(b"\r")):
-          # Non-blank, it's a header
-          if bl is None and L.startswith(b"Content-Length: "):
-            bl=int(L[16:].rstrip())
-          if options.headers:
-            _output(L)
-        else:
-          # Blank line, HTTP header is finished
-          if not options.body:
-            return
-          if options.headers:
-            _output(L)
-          state=2
-          # The above is just for sanity, because we do _not_
-          #  continue with the outer loop,
-          #  since we can now block-output the entire rest of the
-          #  input buffer.
-          if bl is not None:
-            if bl!=wl:
-              print("length mismatch: %s %s %s here: %s given: %s trunc: %s"%\
-                    (length,offset,filename,wl,bl,tr),file=sys.stderr)
-          # HTTP body
-          balance=clear_text.tell()
-          #print(balance,bl,wl,ll,ll-balance,file=sys.stderr)
-          # Output whatever is left
-          _output(cb[balance:balance+wl])
-          return
-
+      # Only output parts (1 = WARC header, 2 = HTTP header, 4 = body) that are wanted
+      bl=None # for HTTP Content-Length for the length of the body?
+      L_start=0
+      state=2
+      with io.BytesIO(bv) as rec_text:
+        for L in rec_text:
+          if state==2:
+            # HTTP header
+            wl -= len(L)
+            if not (L==b"" or L.startswith(b"\r")):
+              # Non-empty, it's (a continuation of) a header
+              if bl is None and L.startswith(b"Content-Length: "):
+                bl=int(L[16:].rstrip())
+            else:
+              # Blank line, HTTP header is finished
+              if parts & 2:
+                callback(wtype,bv[start_2:L_start],2)
+              state=4
+              # The above is just for sanity, because we do _not_
+              #  continue with the outer loop,
+              #  since we can now block-output the entire rest of the
+              #  input buffer.
+              if bl is not None:
+                if bl!=wl:
+                  print("length mismatch: %s %s %s here: %s given: %s trunc: %s"%\
+                        (length,offset,filename,wl,bl,tr),file=sys.stderr)
+              # HTTP body
+              balance=rec_text.tell()
+              #print(balance,bl,wl,ll,ll-balance,file=sys.stderr)
+              # Output whatever is left
+              if parts & 4:
+                callback(wtype,bv[balance:balance+wl],4)
+              state=1
+              
+          L_start=rec_text.tell()
 OUT=open(sys.stdout.fileno(),'wb')
 
 import re
 LMPAT=re.compile(b'^Last-Modified: (.*?)\r',re.MULTILINE)
 
-def showmeLMH(wtype,buf):
+def showmeLMH(wtype,buf,part=2):
   m=LMPAT.search(buf.tobytes(order='A'))
   if m:
     OUT.write(m[1])
   OUT.write(b'\n')
 
-def showme(wtype,buf):
-  OUT.write(buf)
+def showme(wtype,buf,part):
+  if debug:
+    breakpoint()
+  OUT.write(b"%d\n%b"%(part,buf))
 
-warc(showmeLMH,[b'response'])
+#warc(showmeLMH,[b'response'],2)
+
+#warc(showme,[b'response','warcinfo','request','metadata'],int(sys.argv[2]))
+
+warc(showme,[b'response'],int(sys.argv[2]))