misc/scripts/detect_cycle.py
author Sylvain Thénault <sylvain.thenault@logilab.fr>
Wed, 09 Feb 2011 18:06:17 +0100
changeset 6957 ffda12be2e9f
parent 6198 f8dea560703f
child 10589 7c23b7de2b8d
permissions -rw-r--r--
[repository] #1460066: backport datafeed cube as cubicweb source * add some attributes to CWSource to handle this kind of source (not natural to put everything in 'config' string). Adding a CWSource subclass has been attempted then rollbacked because it adds pain to handle multi-sources planning and it introduce an ambiguity on a generic relation (cw_source), which may be a penalty in multiple case * data feed sources are a new kind of source, namely 'copy based', which have no effect on the query planner * a data feed source is associated to a list of url and a parser (appobjects in the 'parsers' registry * entities imported by a data feed have cwuri set to their url on the distant site, their cw_source relation point to the data feed source, though their source stored in the entities table (returned by cw_metainformation) is their physical source, hence 'system'


try:
    rtype, = __args__
except ValueError:
    print 'USAGE: cubicweb-ctl shell <instance> detect_cycle.py -- <relation type>'
    print

graph = {}
for fromeid, toeid in rql('Any X,Y WHERE X %s Y' % rtype):
    graph.setdefault(fromeid, []).append(toeid)

from logilab.common.graph import get_cycles

for cycle in get_cycles(graph):
    print 'cycle', '->'.join(str(n) for n in cycle)