Question

I have a Subversion repository, and I would like to extract detailed information about the history of who edited what and when and how. I know I can run svn log --xml to produce a nice easy-to-use record of which paths were altered in each revision. But what I would also like to find out is the size of the edits to each file.

I know there are many ways to define “edit distance”, but I will be happy with anything simple like “number of lines that are different” for text files.

Presumably I can get all this by parsing the output of svnadmin dump, but then I’d need to spend time learning about the dump file format, which I’d rather avoid if I can.

Answer 1

The svn dump format, it turns out, is very easy to parse. If I generate it with svnadmin dump --deltas then the dump file contains deltas for each file modification, and I can reasonably take the size of the delta (in bytes) to be the edit distance.

In case anyone comes looking here, here is a simple Python script which takes an svn dump file and prints out an XML file containing all the properties. The edit sizes are contained in //path/Text-content-length entries.

def read_defs(f):
    res = {}
    while True:
        l = f.readline()
        if l in ['','\n']: break
        s = l.split(': ',1)
        if len(s)!=2: assert False, 'Bad definition line '+l
        res[s[0]] = (s[1][:-1] if s[1].endswith('\n') else s[1])
    if len(res)==0 and l=='': return None
    return res

def read_props(f):
    res = {}
    lastkey = None
    while True:
        l = f.readline()
        if l.startswith('PROPS-END'): break
        ln = int(l.split()[1])
        l2 = f.read(ln); f.readline()
        if l.startswith('K'):
            lastkey = l2; res[l2] = None
        elif l.startswith('V'):
            res[lastkey] = l2
        else:
            assert False, 'Unexpected prop entry '+l
    return res

def parsedump(f):
        print '<?xml version="1.0"?>'
        print '<log>'
        inrevision,inpaths = False,False
        while True:
            d = read_defs(f)
            if d is None: break
            p = read_props(f) if 'Prop-content-length' in d else {}
            if 'Revision-number' in d:
                if inpaths: print '</paths>'; inpaths=False
                if inrevision: print '</logentry>'
                print '<logentry revision="'+d['Revision-number']+'">'
                inrevision = True
                for k,v in p.iteritems(): print '<'+k+'>'+v+'</'+k+'>'
            elif 'Node-path' in d:
                if not inpaths: print '<paths>'; inpaths=True
                print '<path>'
                for k,v in d.iteritems(): print '<'+k+'>'+v+'</'+k+'>'
                for k,v in p.iteritems(): print '<'+k+'>'+v+'</'+k+'>'
                print '</path>'
            cl = (int(d['Content-length']) if 'Content-length' in d else 0)
            pcl = (int(d['Prop-content-length']) if 'Prop-content-length' in d else 0)
            f.seek(cl-pcl,1)
        if inpaths: print '</paths>'
        if inrevision: print '</logentry>'
        print '</log>'

import sys
if __name__=='__main__':
    if len(sys.argv)==0:
        print 'Usage: svndump2xml FILENAME\nConverts FILENAME to xml, and prints to standard output'
        sys.exit(0)
    filename = sys.argv[1]
    with open(filename,'rb') as f:
        parsedump(f)