#!/bin/env python2 """ gvizify: convert prover9 XML proof files into graphviz ".dot" format (producing a graphical display of a proof); tested with prover9 Aug-2007 Copyright (C) 2007 David A. Wheeler Use '--help' to get information on options. Note: The output of this program is in ".dot" format; you must send it to one of graphviz' programs (usually "dot") to produce real output. This program includes embedded test cases using the 'doctest' format: >>> from gvizify import * """ # WARNING: Some versions of graphviz have a bug that produces bad # Postscript/PDF; if this happens to you, save the generated Postscript # (really EPS), use eps2eps to fix it, and THEN print it or convert to PDF. # Unfortunately, esp2eps produces 'less nice' Postscript. # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2 of the License, or # (at your option) any later version. # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # You should have received a copy of the GNU General Public License # along with this program; if not, write to the Free Software # Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA # Version Information: # 1.3 - Header is now "#!/bin/env python2", adding flexibility # 1.2 - Supports "rotate" for "rotation", add name for entire graph. # 1.1 - Supports environment variable GVIZIFY # 1.0 - includes labels in output by default; --removelabels to prevent # 0.4 - can now control formatting for each node (step) type # 0.3 - add recombine() (reduce lines/node), diff shapes for diff node types # 0.2 - fixed margin problem # 0.1 - initial release # We could use either the "SAX" or "DOM" interface; the DOM interface is # easier to use and more flexible, and the data size of proofs is trivial, # so we'll use DOM. minidom has enough functionality for our needs. from xml.dom.minidom import * import textwrap import sys import os import shlex # Used to process environment variable GVIZIFY from optparse import OptionParser # globals: # options: processed options # args: processed positional arguments # datasource: file or filename to process. def recombine(ls, maxlen): """Given list of strings ls, recombined the (short) strings into longer ones. The strings must not be longer than max(maxlen, longest string in list). >>> recombine(['hi', 'there', 'freddy'], 10) ['hi there', 'freddy'] >>> recombine(['big', 'bug', 'anotherlongone'], 5) ['big bug', 'anotherlongone'] >>> recombine(['reallylongword', 'a', 'b', 'anotherlongone'], 5) ['reallylongword', 'a b', 'anotherlongone'] """ if len(ls) < 2: return ls realmax = max(maxlen, max([len(x) for x in ls])) newls = ls[:] while True: # Loop and 1/2 - anneal list until all recombined changed = False oldls = newls[:] for i in reversed(range(len(oldls))): if i < len(oldls)-1: # If there's a line after line i... newstr = oldls[i] + " " + oldls[i+1] if len(newstr) <= realmax: # If smaller, recombine. newls = oldls[:i] + [newstr] + oldls[i+2:] changed = True break if i > 0: # If there's a line before line i... newstr = oldls[i-1] + " " + oldls[i] if len(newstr) <= realmax: # If smaller, recombine. newls = oldls[:max(0,i-1)] + [newstr] + oldls[i+1:] changed = True break # Done an anneal; either we made a change, or we ran out of things to do. if not changed: break # Completely finished, we've merged all we can. return newls def reformat(s,minwidth=12,maxwidth=17): """Reformat text to multiple lines; we'll try to avoid having tiny lines""" if len(s) <= minwidth: return s bestresult="" bestscore = -100 for curwidth in range(minwidth,maxwidth+1): result = textwrap.wrap(s, width=curwidth, break_long_words=False) result = recombine(result, curwidth) # Recombine if sensible to do so. score = 20-len(result) # More lines penalized score -= 2 * len([x for x in result if len(x)<3]) # short lines penalized score -= 2 * len([x for x in result if len(x)<5]) # if "\n" in result[-1][-2:]: score -= 4 # Really short last line # if "\n" in result[-1][-4:]: score -= 2 # Short last line # print "Score %s for %s" % (score, result) if score>bestscore: bestresult=result bestscore=score # We'll accept the best we found. return "\\n".join(bestresult) def extract_textlist(elements): "Given list of elements e, return as list the text enclosed by each." resultfragments = [] for element in elements: textfragment = [] for child in element.childNodes: if child.nodeType in (child.TEXT_NODE, child.CDATA_SECTION_NODE): textfragment.append(child.data) resultfragments.append( "".join(textfragment) ) return resultfragments def process_command_line(): global options, args, datasource # This function sets these. parser = OptionParser( usage = "usage: %prog [options] [XML filename]", version="%prog 1.3, 2007-11-03", description= """This program by David A. Wheeler translates prover9 XML data into graphviz format, so that proofs are much easier to follow. Parameters may be optionally included in the environment variable GVIZIFY, e.g: GVIZIFY='--rotation=0 --relax --size=42,42' gvizify stuff.xml > stuff.dot """ ) parser.add_option("-u", "--unnumbered", action="store_false", dest="numberit", default=True, help="do not number each step") parser.add_option("--minwidth", dest="minwidth", default=12, metavar="WIDTH", help="minimum width (less than this stays on one line)") parser.add_option("--maxwidth", dest="maxwidth", default=17, metavar="WIDTH", help="minimum width ('longest' allowed line)") parser.add_option("--removelabels", action="store_false", dest="labels", default=True, help="remove labels (normally they're included)") parser.add_option("--rotation", "--rotate", dest="rotation", default=90, metavar="ANGLE", help="rotation (in degrees; 90 is landscape)") parser.add_option("--size", dest="size", default="10,7.5", help="papersize 'width,length' in inches minus margins") parser.add_option("--multipage", dest="multipage", default=False, action="store_true", help="spread to multiple pages (use --size also)") parser.add_option("--relax", action="store_true", dest="relax", default=False, help="use 'natural' graph size (vs. expanding to page size)") parser.add_option("--margin", dest="margin", default="0.5,0.5", help="paper margin 'width,length' in inches") parser.add_option("--command", dest="command", default="", help="literal graphviz command (inserted into output)") parser.add_option("--defaultstep", dest="defaultstep", default='shape="ellipse"', metavar="GVIZATTRIBUTE", help="node attribute for ordinary steps (graphviz format)") parser.add_option("--assumption", dest="assumption", default='shape="box"', metavar="GVIZATTRIBUTE", help="node attribute for assumption steps (graphviz format)") parser.add_option("--clausify", dest="clausify", default='shape="plaintext"', metavar="GVIZATTRIBUTE", help="node attribute for clausify steps (graphviz format)") parser.add_option("--goal", dest="goal", default=' shape="doubleoctagon"', metavar="GVIZATTRIBUTE", help="node attribute for goal steps (graphviz format)") parser.add_option("--deny", dest="deny", default=' shape="octagon"', metavar="GVIZATTRIBUTE", help="node attribute for deny steps (graphviz format)") parser.add_option("--name", dest="name", default="", help="name of entire graph") parser.add_option("--selftest", action="store_true", dest="selftest", default=False, help="run trivial self-test; no output==okay") optionparams = sys.argv[1:] if 'GVIZIFY' in os.environ: optionparams += shlex.split(os.environ['GVIZIFY']) (options, args) = parser.parse_args(args=optionparams) if len(args) == 0: datasource=sys.stdin if options.name == "": options.name = "-" elif len(args) == 1: datasource=args[0] if options.name == "": options.name = args[0].replace(".xml","") else: print "Sorry, must provide at most 1 argument (the XML's filename)" exit(1) if options.name == "": options.name = "Graph" def generate_results(): "Parse XML from datasource, and write to stdout the graphviz results" # Generate header: print "strict digraph \"%s\" {" % (options.name.replace('"', r'\"') ,) print " rotate=%s;" % (options.rotation,) if options.multipage: print " page=\"%s\";" % (options.size,) else: force="" if options.relax else "!" print " size=\"%s%s\";" % (options.size, force) print " margin=\"%s\";" % (options.margin,) if options.command != "": print options.command # Process XML file: doc = parse(datasource) # [0] is the program name. prooflist = doc.getElementsByTagName("proof") if (prooflist.length != 1): print "Error! Must be exactly 1 proof item in the XML file" exit(1) proof=prooflist.item(0) for clause in proof.getElementsByTagName("clause"): # At a clause, get its id. step = clause.getAttribute("id") # Get and format its result literals = clause.getElementsByTagName("literal") literaltexts = extract_textlist(literals) # If >1 literal, they're implicitly joined by "|": result = " | ".join(literaltexts) if options.labels: attributes = clause.getElementsByTagName("attribute") attributetexts = extract_textlist(attributes) # Keep only labels, and remove label(non_clause), label(goal) attributetexts = [x for x in attributetexts if (("label(" in x) and ("label(non_clause)" not in x) and ("label(goal)" not in x))] attributetexts = [" # " + x for x in attributetexts] attributeresults = " ".join(attributetexts) result += attributeresults result = result.replace("\n"," ").strip() result = " ".join(result.split()) # Remove duplicate whitespace if options.numberit: result = "{%s} %s" % (step, result) result = reformat(result,options.minwidth,options.maxwidth) # Determine parent list (assumptions won't have any) parentlist = [] justification = clause.getElementsByTagName("justification")[0] for rule in justification.childNodes: if rule.nodeType == rule.ELEMENT_NODE: parentlist.extend([int(x) for x in rule.getAttribute("parents").split()]) steptype = clause.getAttribute("type") if steptype in ('assumption', 'clausify', 'goal', 'deny'): extras=',' + getattr(options,steptype) else: extras=',' + options.defaultstep print " %s [label=\"%s\"%s];" % (step, result, extras) for edge in parentlist: print " %s -> %s;" % (edge, step) # Done with the graph; generate graph footer. print " root=%s; // last step is plausible root" % step print "}" def test(): import doctest doctest.testmod() def main(): process_command_line() if options.selftest: test() else: generate_results() if __name__ == "__main__": main()