mirror of
https://github.com/safishamsi/graphify.git
synced 2026-10-14 16:09:26 +00:00
devkancheti4-design
d3f8687c00
_pick_seeds appended the per-term guaranteed seeds (#1445) in sorted(best_seed_by_term), alphabetical term order,
and the renderer prints seeds first, so a weak match on an early-alphabet term led a specific term's exact match.
_score_query now returns term_order: the per-term winners by their existing key (-singleton score, which carries
the term's IDF weight, then -degree, label length, node id), the term breaking ties. _query_graph_text passes it to
_pick_seeds. The same seeds are picked and the order is deterministic; without term_order the sorted order is kept.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
(cherry picked from commit 7bcbc60ac1)
74 lines
3.7 KiB
Python
74 lines
3.7 KiB
Python
"""The per-term guaranteed seeds follow the strength of each term's best match, strongest first.
|
|
|
|
`_pick_seeds` guarantees one seed per query term (#1445). Those seeds used to be appended in alphabetical term
|
|
order, and the renderer prints seeds first, so on a long query a weak match on an early-alphabet term was printed
|
|
ahead of a specific term's exact match. `_score_query` now returns `term_order` (strongest per-term winner first,
|
|
the term breaking ties) and `_query_graph_text` passes it on. Which seeds are picked does not change, only their
|
|
order, and the order stays deterministic."""
|
|
from __future__ import annotations
|
|
|
|
import networkx as nx
|
|
|
|
from graphify.serve import _pick_seeds, _query_graph_text, _score_query
|
|
|
|
|
|
def _graph():
|
|
G = nx.Graph()
|
|
G.add_node("top", label="TopLevelRouter", source_file="router.py", source_location="L1")
|
|
G.add_node("apple_weak", label="apple_pie_formatting_helper", source_file="fmt.py", source_location="L3")
|
|
G.add_node("zebra_exact", label="zebra", source_file="zebra.py", source_location="L1")
|
|
for i in range(6): # make "apple" common, so its match weighs less
|
|
G.add_node(f"apple{i}", label=f"apple_util_{i}", source_file=f"apples{i}.py", source_location="L1")
|
|
G.add_edge("top", "apple_weak")
|
|
G.add_edge("top", "zebra_exact")
|
|
return G
|
|
|
|
|
|
def test_pick_seeds_follows_term_order_when_given():
|
|
G = _graph()
|
|
scored = [(1000.0, "top"), (1.0, "apple_weak"), (1.0, "zebra_exact")]
|
|
best = {"apple": "apple_weak", "zebra": "zebra_exact"}
|
|
assert _pick_seeds(scored, G=G, best_seed_by_term=best, term_order=("zebra", "apple")) == ["top", "zebra_exact", "apple_weak"]
|
|
|
|
|
|
def test_pick_seeds_keeps_sorted_term_order_without_it():
|
|
G = _graph()
|
|
scored = [(1000.0, "top"), (1.0, "apple_weak"), (1.0, "zebra_exact")]
|
|
best = {"apple": "apple_weak", "zebra": "zebra_exact"}
|
|
assert _pick_seeds(scored, G=G, best_seed_by_term=best) == ["top", "apple_weak", "zebra_exact"]
|
|
|
|
|
|
def test_pick_seeds_appends_terms_missing_from_term_order_in_sorted_order():
|
|
G = _graph()
|
|
scored = [(1000.0, "top")]
|
|
best = {"apple": "apple_weak", "zebra": "zebra_exact"}
|
|
assert _pick_seeds(scored, G=G, best_seed_by_term=best, term_order=("zebra",)) == ["top", "zebra_exact", "apple_weak"]
|
|
|
|
|
|
def test_score_query_orders_terms_by_their_best_match_strength():
|
|
qs = _score_query(_graph(), ["apple", "zebra"], collect_per_term_seeds=True)
|
|
assert set(qs.term_order) == set(qs.best_seed_by_term)
|
|
assert qs.term_order[0] == "zebra" # an exact label match on a rare term leads
|
|
assert qs.best_seed_by_term["zebra"] == "zebra_exact"
|
|
|
|
|
|
def _five_term_graph():
|
|
"""Five terms with exact label matches; "apple" is common (eight more labels contain it), so its exact match
|
|
carries less IDF weight than the rare "charlie". Three seeds fill the gap window; "apple" and "charlie" arrive
|
|
through the per-term guarantee."""
|
|
G = nx.Graph()
|
|
for name in ("alpha", "bravo", "charlie", "zebra", "apple"):
|
|
G.add_node(name, label=name, source_file=f"{name}.py", source_location="L1")
|
|
for i in range(8):
|
|
G.add_node(f"apple{i}", label=f"apple_util_{i}", source_file=f"apples{i}.py", source_location="L1")
|
|
return G
|
|
|
|
|
|
def test_query_prints_the_stronger_guaranteed_seed_first_and_is_deterministic():
|
|
G = _five_term_graph()
|
|
text = _query_graph_text(G, "alpha bravo charlie apple zebra", mode="bfs", depth=1)
|
|
assert text == _query_graph_text(G, "alpha bravo charlie apple zebra", mode="bfs", depth=1)
|
|
start = text.split("Start: ", 1)[1].split("]", 1)[0]
|
|
# Alphabetical term order printed the common term's seed ("apple") ahead of the rare exact match ("charlie").
|
|
assert start.index("'charlie'") < start.index("'apple'"), start
|