def parse_workflow():
# FIXME TODO these states should probably be compiled down to numbers???
docs = Path(__file__).parent.absolute().resolve().parent / 'docs'
rridpath = docs / 'workflow-rrid.graphml'
paperpath = docs / 'workflow-paper-id.graphml'
cgraph = ConjunctiveGraph()
gt.WorkflowMapping(rridpath.as_posix()).graph(cgraph)
gt.PaperIdMapping(paperpath.as_posix(), False).graph(cgraph)
write(cgraph, '/tmp/workflow.ttl')
predicates = set(cgraph.predicates())
OntCuries({cp:str(ip) for cp, ip in cgraph.namespaces()})
OntCuries({'RRID': 'https://scicrunch.org/resolver/RRID:',
'DOI': 'https://doi.org/',
'PMID': 'https://www.ncbi.nlm.nih.gov/pubmed/'})
hg = makeGraph('', graph=cgraph)
short = sorted(hg.qname(_) for _ in predicates)
wf.hasTag
wf.hasReplyTag
wf.hasTagOrReplyTag
wf.hasOutputTag
#if type isa wf.tag
tag_types = set(cgraph.transitive_subjects(rdfs.subClassOf, wf.tag))
tag_tokens = {tagType:sorted(set(t for t in cgraph.transitive_subjects(rdf.type, tagType)
if t != tagType))
for tagType in tag_types}
has_tag_types = set(cgraph.transitive_subjects(rdfs.subPropertyOf, wf.hasTagOrReplyTag))
has_tag_types.add(wf.hasOutputTag)
has_next_action_types = set(cgraph.transitive_subjects(rdfs.subPropertyOf, wf.hasOutput))
has_next_action_types.add(wf.hasNextStep)
terminals = sorted(tag
for ttype in tag_types
if ttype != wf.tagScibot # scibot is not 'terminal' for this part
for tag in cgraph[:rdf.type:ttype]
if not isinstance(tag, BNode)
and not any(o for httype in has_tag_types
for o in cgraph[tag:httype]))
endpoints = sorted(endpoint
for endpoint in cgraph[:rdf.type:wf.state]
if not isinstance(endpoint, BNode)
and not any(o for hnatype in has_next_action_types
for o in cgraph[endpoint:hnatype]))
complicated = sorted(a_given_tag
for tt in tag_types
for a_given_tag in cgraph[:rdf.type:tt]
if not isinstance(a_given_tag, BNode)
and not [successor_tag
for htt in has_tag_types
for successor_tag in chain(t
for t in cgraph[a_given_tag:htt]
#if not isinstance(t, BNode)
,
# we don't actually need this for terminals
# we will need it later
#(t for b in cgraph[a_given_tag:htt]
#if isinstance(b, BNode)
#for listhead in cgraph[b:owl.oneOf]
#for t in unlist(listhead, cgraph)),
)])
def topList(node, g):
for s in g[:rdf.rest:node]:
yield s
def getLists(node, g):
for linker in g[:rdf.first:node]:
top = None
for top in g.transitiveClosure(topList, linker):
pass
if top:
yield top
else:
yield linker
def getIsTagOf(node, g):
for htt in has_tag_types:
for parent_tag in g[:htt:node]:
yield parent_tag
def getIsOneOfTagOf(node, g):
for list_top in getLists(node, g):
for linker in g[:owl.oneOf:list_top]:
for parent_tag, _ in g[::linker]:
yield parent_tag
def getPreviousTag(node, g): # not quite what we need
yield from getIsOneOfTagOf(node, g)
yield from getIsTagOf(node, g)
def getTagChains(node, g, seen=tuple()):
# seen to prevent recursion cases where
# taggning can occur in either order e.g. PMID -> DOI
#print(tc.red(repr(OntId(node)))) # tc.red(OntId(node)) does weird stuff O_o
parent_tag = None
for parent_tag in chain(getIsOneOfTagOf(node, g),
getIsTagOf(node, g)):
if parent_tag in seen:
parent_tag = None
continue
ptt = next(g[parent_tag:rdf.type])
#if ptt in tag_types:
for pchain in getTagChains(parent_tag, g, seen + (node,)):
if ptt in tag_types:
out = parent_tag, *pchain
else:
out = pchain
yield out
if not ptt and not out:
parent_tag = None
if not parent_tag:
yield tuple()
def getInitiatesAction(node, g):
for action in g[:wf.initiatesAction:node]:
yield action
def getIsOneOfOutputOf(node, g):
for list_top in getLists(node, g):
for linker in g[:owl.oneOf:list_top]:
for hot in has_next_action_types:
for parent_thing in g[:hot:linker]:
yield parent_thing
def getActionChains(node, g):
parent_action = None
for parent_action in chain(getIsOneOfOutputOf(node, g), # works for actions too
getInitiatesAction(node, g)):
for pchain in getActionChains(parent_action, g): # NOTE may also be a tag...
out = parent_action, *pchain
#print(tuple(hg.qname(o) for o in out))
yield out
if not parent_action:
yield tuple()
def getRestSubjects(predicate, object, g):
""" invert restriction """
rsco = cmb.Restriction(rdfs.subClassOf)
for rt in rsco.parse(graph=g):
if rt.p == predicate and rt.o == object:
yield from g.transitive_subjects(rdfs.subClassOf, rt.s)
annoParts = list(getRestSubjects(wf.isAttachedTo, wf.annotation, cgraph))
partInstances = {OntId(a):set(t if isinstance(t, BNode) else OntId(t)
for t in cgraph.transitive_subjects(rdf.type, a)
if not isinstance(t, BNode) and t != a)
for a in annoParts}
_endpoint_chains = {OntId(endpoint):[[OntId(endpoint)] + [OntId(e) for e in chain]
for chain in getActionChains(endpoint, cgraph)]
for endpoint in endpoints}
#print([hg.qname(e) for e in endpoints])
#print([print([hg.qname(c) for c in getActionChains(endpoint, cgraph) if c])
#for endpoint in endpoints
#if endpoint])
#_ = [print(list(getActionChains(e, cgraph)) for e in endpoints)]
#return
wat = cgraph.transitiveClosure(getPreviousTag, RRIDCUR.Duplicate)
wat = list(wat)
#def invOneOf(tag, g):
fake_chains = {hg.qname(terminal):
[hg.qname(c)
for c in cgraph.transitiveClosure(getPreviousTag, terminal)]
for terminal in terminals}
def make_chains(things, getChains):
return {OntId(thing):[[OntId(thing)] + [OntId(e) for e in chain]
for chain in getChains(thing, cgraph)]
for thing in things
#if not print(thing)
}
def print_chains(thing_chains):
print('\nstart from beginning')
print('\n'.join(sorted(' -> '.join(hg.qname(e) for e in reversed(chain))
for chains in thing_chains.values()
for chain in chains)))
print('\nstart from end')
print('\n'.join(sorted(' <- '.join(e.curie for e in chain)
for chains in thing_chains.values()
for chain in chains)))
def valid_tagsets(all_chains):
# not the most efficient way to do this ...
transitions = defaultdict(set)
for end, chains in all_chains.items():
for chain in chains:
valid = set()
prior_state = None
for element in reversed(chain):
valid.add(element)
state = frozenset(valid)
transitions[prior_state].add(state)
prior_state = state
return {s:frozenset(n) for s, n in transitions.items()}
endpoint_chains = make_chains(endpoints, getActionChains)
#endpoint_transitions = valid_transitions(endpoint_chains) # not the right structure
print_chains(endpoint_chains)
terminal_chains = make_chains(terminals, getTagChains)
print_chains(terminal_chains)
tag_transitions = valid_tagsets(terminal_chains)
terminal_tags_to_endpoints = 'TODO'
def printq(*things):
print(*(OntId(t).curie for t in things))
from pprint import pprint
def get_linkers(s, o, g, linkerFunc): # FIXME not right
for p in g[s::o]:
yield p
for l in linkerFunc(o, g):
#print(tc.blue(f'{OntId(s).curie} {l if isinstance(l, BNode) else OntId(l).curie}'))
for p in g[s::l]:
#print(tc.red(f'{s} {l} {o} {p}'))
yield p
return
linkers = set(l for l in g.transitiveClosure(linkerFunc, o))
for p, o in g[s::]:
if o in linkers:
yield p
def edge_to_symbol(p, rev=False):
if p == wf.initiatesAction:
return '<<' if rev else '>>'
elif p == wf.hasReplyTag:
return '<' if rev else '>'
elif p == wf.hasTagOrReplyTag:
return '<=' if rev else '=>'
elif p == wf.hasOutputTag:
return '-<-' if rev else '->-'
else:
return '<??' if rev else '??>'
def chain_to_typed_chain(chain, g, func):
# duh...
#pprint(chain)
for s, o in zip(chain, chain[1:]):
# TODO deal with reversed case
s, o = s.u, o.u
p = None
#print(s, o)
printq(s, o)
for p in get_linkers(s, o, g, func):
#print(tc.yellow(p))
#yield (s, edge_to_symbol(p), o)
yield from (s, edge_to_symbol(p), o)
if not p:
for rp in get_linkers(o, s, g, func):
print(tc.blue(rp))
yield from (s, edge_to_symbol(rp, rev=True), o)
def tchains(thing_chains, func):
return sorted([OntId(e).curie if isinstance(e, URIRef) else e
for e in chain_to_typed_chain(list(reversed(chain)), cgraph, func)]
for chains in thing_chains.values()
for chain in chains)
def getLinkers(node, g):
for list_top in getLists(node, g):
for linker in g[:owl.oneOf:list_top]:
yield linker
def allSubjects(object, graph):
yield from (s for s, p in graph[::object])
yield from getLinkers(object, graph)
print()
ttc = tchains(terminal_chains, allSubjects)
tec = tchains(endpoint_chains, allSubjects)
pprint(ttc)
pprint(tec)
valid_tagsets = frozenset((t for s in tag_transitions.values() for t in s))
tts = valid_tagsets - frozenset(tag_transitions)
endtype = 'TODO' #
tt = {}
for endtype, chains in endpoint_chains.items():
for *_chain, tag in chains:
if _chain:
next_thing = _chain[-1]
for ets in tts:
if tag in ets:
tt[ets] = next_thing
terminal_tagsets = tt
#[print(wat) for wat in terminal_chains.values()]
#pprint(terminal_chains)
return tag_types, tag_tokens, partInstances, valid_tagsets, terminal_tagsets, tag_transitions
