Skip to content

Codebase intelligence for AI and humans: code health scores, auto-generated docs, git analytics, dead code detection, and architectural decisions via MCP.

fix(health): one perf plan step per call site, naming the call it batches (#3250)

Raghav Chamadiya323fc636645d
diff --git a/docs/architecture/code-health.md b/docs/architecture/code-health.md
index 4f1fb9d6e..ab0a43021 100644
--- a/docs/architecture/code-health.md
+++ b/docs/architecture/code-health.md
@@ -716,10 +716,13 @@ by `rank_score`; actionability only breaks ties. Excluded reasons are counted in
against a DB or network client is advisory with a `bounded_concurrency`
prerequisite. A loop already walking chunks is `loop_already_chunked`. When
`io_in_loop` and `serial_await_in_loop` fire on one call they name each other in
-`siblings`. Plans list their edits (`mechanical` only when proven) and validation
-tests (coverage, call graph, import graph, then `via: "name-match"`). An
-opportunity links to a `performance_fix` refactoring plan only by exact
-`opportunity_id`.
+`siblings`. Plans list their edits (`mechanical` only when proven): one step per
+call site and loop, however many markers saw it, each naming the call it repeats
+(`details.sink_call`, or the helper a cross-function path enters). A step that adds
+a bulk form to the sink comes first only when every site reaches that one sink.
+Plans also list validation tests (coverage, call graph, import graph, then
+`via: "name-match"`). An opportunity links to a `performance_fix` refactoring
+plan only by exact `opportunity_id`.
**Dataflow promotion.** `serial_await_in_loop` and `nested_loop_quadratic` are
promoted to `dataflow_verified: true` when a def/use pass proves iterations
diff --git a/packages/core/src/repowise/core/analysis/health/complexity/models.py b/packages/core/src/repowise/core/analysis/health/complexity/models.py
index f661bea75..6eacb2005 100644
--- a/packages/core/src/repowise/core/analysis/health/complexity/models.py
+++ b/packages/core/src/repowise/core/analysis/health/complexity/models.py
@@ -316,12 +316,20 @@ class PerfHit:
# Distinct direct callers of the enclosing function, set only on the
# centrality-gated ``hot_path_sync_io`` hit; 0 everywhere else.
callers: int = 0
+ # The per-iteration call as written at ``line`` (``session.get``), so a plan
+ # can name what to batch; empty where the detector kept no call text.
+ sink_call: str = ""
def loop_facts(self) -> dict[str, Any]:
- """Loop facts for ``details``; absent when unset so old findings are unchanged."""
+ """Loop facts and the call text for ``details``.
+
+ Each key is absent when unset, so old findings are unchanged.
+ """
facts = self.loop.as_details() if self.loop is not None else {}
if self.loop_line:
facts["loop_line"] = self.loop_line
+ if self.sink_call:
+ facts["sink_call"] = self.sink_call
return facts
diff --git a/packages/core/src/repowise/core/analysis/health/complexity/perf_walk.py b/packages/core/src/repowise/core/analysis/health/complexity/perf_walk.py
index 78246336a..126ae0ef6 100644
--- a/packages/core/src/repowise/core/analysis/health/complexity/perf_walk.py
+++ b/packages/core/src/repowise/core/analysis/health/complexity/perf_walk.py
@@ -327,6 +327,48 @@ def _loop_key_feeds(call: Node, loop: Node) -> bool:
return bool(_loop_bound_names(loop) & _identifiers(call))
+_SINK_CALL_MAX = 60
+
+
+_STRING_QUOTES = frozenset("'\"`")
+
+
+def _sink_call_text(call: Node, method: str, dialect: BasePerfDialect) -> str:
+ """The callee as written with its arguments dropped, else its method name.
+
+ ``session.get`` stays as is and ``conn.execute(sql).fetchone`` reads
+ ``conn.execute().fetchone``. A callee that opens with a parenthesis, runs
+ long or holds a string literal (its quotes defeat the paren scan) gives the
+ method only, as does a dialect with no callee text.
+ """
+ text = dialect.callee_text(call)
+ if len(text) > 4 * _SINK_CALL_MAX or not _STRING_QUOTES.isdisjoint(text):
+ return method
+ callee = _strip_call_args(text)
+ return callee if _quotable(callee) else method
+
+
+def _strip_call_args(text: str) -> str:
+ """``conn.execute(sql).fetchone`` -> ``conn.execute().fetchone``, whitespace dropped."""
+ if "(" not in text:
+ return "".join(text.split())
+ kept: list[str] = []
+ depth = 0
+ for char in text:
+ if char == "(":
+ kept.append("()" if depth == 0 else "")
+ depth += 1
+ elif char == ")":
+ depth = max(depth - 1, 0)
+ elif depth == 0 and not char.isspace():
+ kept.append(char)
+ return "".join(kept)
+
+
+def _quotable(callee: str) -> bool:
+ return bool(callee) and not callee.startswith("(") and len(callee) <= _SINK_CALL_MAX
+
+
def _key_unused(hit: PerfHit) -> bool:
return hit.loop is not None and hit.loop.key_unused
@@ -634,8 +676,12 @@ def _collect_perf_hits(
if kind is not None:
if loop_depth >= 1:
facts = loop_facts(call_node, sink=True, per_call=True)
+ call = _sink_call_text(call_node, method, dialect)
hits.append(
- PerfHit("io_in_loop", line, next_func, kind, func_start=next_start, loop=facts)
+ PerfHit(
+ "io_in_loop", line, next_func, kind,
+ func_start=next_start, loop=facts, sink_call=call,
+ )
)
if do_serial_await and awaited:
# An *awaited* sink in a loop body is additionally a
@@ -646,7 +692,7 @@ def _collect_perf_hits(
hits.append(
PerfHit(
"serial_await_in_loop", line, next_func, kind,
- func_start=next_start, loop=facts,
+ func_start=next_start, loop=facts, sink_call=call,
)
)
if do_nested_io and loop_depth >= 2 and outer_iter:
@@ -660,7 +706,7 @@ def _collect_perf_hits(
hits.append(
PerfHit(
"nested_loop_with_io", line, next_func, kind,
- func_start=next_start, loop=facts,
+ func_start=next_start, loop=facts, sink_call=call,
)
)
else:
diff --git a/packages/core/src/repowise/core/analysis/health/perf/dialects/base.py b/packages/core/src/repowise/core/analysis/health/perf/dialects/base.py
index a8f1a4967..c750bf6e7 100644
--- a/packages/core/src/repowise/core/analysis/health/perf/dialects/base.py
+++ b/packages/core/src/repowise/core/analysis/health/perf/dialects/base.py
@@ -155,6 +155,11 @@ class BasePerfDialect:
txt = (node.text or b"").decode("utf-8", "replace")
return txt.split(".")[0] if txt else None
+ def callee_text(self, call_node: Node) -> str:
+ """The callee as written, arguments included (``session.get``); empty if unknown."""
+ fn = call_node.child_by_field_name("function")
+ return (fn.text or b"").decode("utf-8", "replace") if fn is not None else ""
+
def callee_method_name(self, call_node: Node) -> str | None:
"""Rightmost member of the callee (``x.execute`` -> 'execute')."""
fn = call_node.child_by_field_name("function")
diff --git a/packages/core/src/repowise/core/analysis/health/perf/dialects/java.py b/packages/core/src/repowise/core/analysis/health/perf/dialects/java.py
index 65d814b13..d4d58c691 100644
--- a/packages/core/src/repowise/core/analysis/health/perf/dialects/java.py
+++ b/packages/core/src/repowise/core/analysis/health/perf/dialects/java.py
@@ -260,6 +260,15 @@ class JavaPerfDialect(BasePerfDialect):
return name.text.decode("utf-8", "replace")
return None
+ def callee_text(self, call_node: Node) -> str:
+ # ``method_invocation`` splits the callee into ``object`` and ``name``.
+ name = call_node.child_by_field_name("name")
+ if call_node.type != "method_invocation" or name is None or not name.text:
+ return ""
+ obj = call_node.child_by_field_name("object")
+ method = name.text.decode("utf-8", "replace")
+ return f"{obj.text.decode('utf-8', 'replace')}.{method}" if obj and obj.text else method
+
def callee_root_name(self, call_node: Node) -> str | None:
if call_node.type == "object_creation_expression":
t = call_node.child_by_field_name("type")
diff --git a/packages/core/src/repowise/core/analysis/health/perf/dialects/kotlin.py b/packages/core/src/repowise/core/analysis/health/perf/dialects/kotlin.py
index 2794df692..f1d2c0984 100644
--- a/packages/core/src/repowise/core/analysis/health/perf/dialects/kotlin.py
+++ b/packages/core/src/repowise/core/analysis/health/perf/dialects/kotlin.py
@@ -239,6 +239,9 @@ class KotlinPerfDialect(BasePerfDialect):
return _decode(fn)
return None
+ def callee_text(self, call_node: Node) -> str:
+ return _decode(_callee(call_node)) or ""
+
def callee_root_name(self, call_node: Node) -> str | None:
node = _callee(call_node)
# Walk to the bottom of a ``a.b().c()`` chain — the leftmost receiver is
diff --git a/packages/core/src/repowise/core/analysis/health/perf/facts.py b/packages/core/src/repowise/core/analysis/health/perf/facts.py
index 83c998b93..f4f3e9309 100644
--- a/packages/core/src/repowise/core/analysis/health/perf/facts.py
+++ b/packages/core/src/repowise/core/analysis/health/perf/facts.py
@@ -150,6 +150,7 @@ def evidence_row(facts: ObservationFacts) -> dict[str, Any]:
# The loop around the call, where a reader starts; absent when the
# walk stored none, so older evidence reads unchanged.
**({"loop_line": loop} if (loop := facts.details.get("loop_line")) else {}),
+ **({"sink_call": call} if (call := facts.details.get("sink_call")) else {}),
}
diff --git a/packages/core/src/repowise/core/analysis/health/refactoring/performance_fix.py b/packages/core/src/repowise/core/analysis/health/refactoring/performance_fix.py
index cfe074933..f7b96068e 100644
--- a/packages/core/src/repowise/core/analysis/health/refactoring/performance_fix.py
+++ b/packages/core/src/repowise/core/analysis/health/refactoring/performance_fix.py
@@ -4,6 +4,7 @@ from __future__ import annotations
from collections.abc import Iterable
from dataclasses import dataclass
+from typing import Any
from repowise.core.analysis.execution_graph import file_of_symbol
@@ -14,15 +15,16 @@ from .models import RefactoringSuggestion
from .preconditions import Applicability
# Per strategy: an optional step at the intervention, then one step per call
-# site. Templates name the edit; the locations are the evidence's own.
+# site. Templates name the edit; the locations are the evidence's own, and
+# ``{call}`` names the call each site repeats.
_STEPS: dict[str, tuple[str | None, str]] = {
"batch_or_prefetch_io": (
"Add a form of {sink} that takes every key at once",
- "Collect the keys before the loop and call the batched form once",
+ "Collect the keys before the loop and replace the per-key {call} with one batched call",
),
"parallelize_independent_awaits": (
None,
- "Run the independent awaits concurrently, with a bound on how many at once",
+ "Run the independent {call}s concurrently, with a bound on how many at once",
),
"replace_membership_collection": (None, "Build a set once before the loop and probe it"),
"buffer_string_accumulation": (None, "Accumulate into a list and join once after the loop"),
@@ -37,11 +39,11 @@ _API_STEPS: dict[str, tuple[str | None, str]] = {
"batch_or_prefetch_io": (
None,
"Collect the keys before the loop and make one call with {api} in place of the "
- "per-key call",
+ "per-key {call}",
),
"parallelize_independent_awaits": (
None,
- "Run the independent awaits concurrently, each still inside {api}",
+ "Run the independent {call}s concurrently, each still inside {api}",
),
}
@@ -67,13 +69,18 @@ def fix_steps(
locations: list[dict[str, object]],
api: str | None = None,
) -> list[dict[str, object]]:
- """Ordered edits for one plan; mechanical only when the strategy is proven."""
+ """Ordered edits for one plan; mechanical only when the strategy is proven.
+
+ *sink* is the sink every site shares, or ``None``: a bulk form added to one
+ member's sink would leave the other sites' calls per-key.
+ """
intro, per_site = (_API_STEPS if api else _STEPS).get(strategy, (None, ""))
- per_site = per_site.format(api=api) if api else per_site
applicability: Applicability = "mechanical" if safety == "proven" else "judgment"
steps: list[dict[str, object]] = []
- if intro:
- target = intervention or sink
+ target = intervention or sink
+ # With nothing to name, a site step says the whole edit; a strategy with no
+ # site step keeps its intro whatever it can name.
+ if intro and (target or not per_site):
steps.append(
{
"action": intro.format(sink=(target or "the per-key call").rsplit("::", 1)[-1]),
@@ -86,7 +93,7 @@ def fix_steps(
if per_site:
steps.extend(
{
- "action": per_site,
+ "action": per_site.format(api=api, call=_call_phrase(location.get("call"))),
"symbol": location.get("function_name"),
"file_path": location.get("file_path"),
"line": location.get("line_start"),
@@ -98,6 +105,10 @@ def fix_steps(
return [{"order": index, **step} for index, step in enumerate(steps, 1)]
+def _call_phrase(call: object) -> str:
+ return f"{call} call" if call else "call"
+
+
_EFFORT_BANDS = ((1, "S"), (3, "M"), (8, "L"))
@@ -139,6 +150,54 @@ def _intervention(opportunity: PerformanceOpportunity, fix: PerformanceFix) -> s
return opportunity.intervention_symbol
+def _site_locations(evidence: Iterable[dict[str, Any]]) -> list[dict[str, Any]]:
+ """One location per call site and loop, whatever number of markers observed it.
+
+ ``io_in_loop`` and ``nested_loop_with_io`` both fire on one call; that is one
+ edit, not two.
+ """
+ sites: dict[tuple[Any, Any, Any], dict[str, Any]] = {}
+ for item in evidence:
+ key = (item["file_path"], item.get("line_start"), item.get("loop_line"))
+ site = sites.setdefault(
+ key,
+ {
+ "file_path": item["file_path"],
+ "function_name": item.get("function_name"),
+ "line_start": item.get("line_start"),
+ "line_end": item.get("line_end"),
+ **({"loop_line": item["loop_line"]} if item.get("loop_line") else {}),
+ },
+ )
+ if "call" not in site and (call := _site_call(item)):
+ site["call"] = call
+ return list(sites.values())
+
+
+def _site_call(item: dict[str, Any]) -> str | None:
+ """The call the site repeats: the detector's text, else the helper hop.
+
+ A cross-function path starts at the loop's own function (``path[0]``), so
+ ``path[1]`` is the helper the loop calls at this site.
+ """
+ path = item.get("path") or ()
+ hop = path[1] if len(path) > 1 and isinstance(path[1], str) else None
+ return item.get("sink_call") or (hop.rsplit("::", 1)[-1] if hop else None)
+
+
+def _shared_sink(opportunity: PerformanceOpportunity) -> str | None:
+ """The sink every site reaches, or ``None`` when a site has none or they differ.
+
+ Evidence is capped; past the cap only the members' resolved sinks are known,
+ so a pathless member beyond it goes unseen (ceiling: carry a has-path count).
+ """
+ sinks = {item["path"][-1] if item.get("path") else None for item in opportunity.evidence}
+ sink = sinks.pop() if len(sinks) == 1 else None
+ if opportunity.evidence_truncated and sink != opportunity.terminal_sink:
+ return None
+ return sink
+
+
def _suggestion(
opportunity: PerformanceOpportunity, fix: PerformanceFix, confidence: str
) -> RefactoringSuggestion:
@@ -146,16 +205,7 @@ def _suggestion(
intervention = _intervention(opportunity, fix)
target_file = file_of_symbol(intervention) if intervention else anchor["file_path"]
in_anchor_file = target_file == anchor["file_path"]
- locations = [
- {
- "file_path": item["file_path"],
- "function_name": item.get("function_name"),
- "line_start": item.get("line_start"),
- "line_end": item.get("line_end"),
- **({"loop_line": item["loop_line"]} if item.get("loop_line") else {}),
- }
- for item in opportunity.evidence
- ]
+ locations = _site_locations(opportunity.evidence)
# A bulk form belongs on the per-key callee: the shared helper, else the one
# sink the loop reaches. The loop's own function is where the keys are
# collected, never what gains the bulk form.
@@ -164,7 +214,7 @@ def _suggestion(
fix.strategy,
fix.safety,
intervention if fix.strategy != "batch_or_prefetch_io" or batched_on_helper else None,
- opportunity.terminal_sink,
+ _shared_sink(opportunity),
locations,
fix.api,
)
diff --git a/tests/fixtures/perf_corpus/golden_plans.json b/tests/fixtures/perf_corpus/golden_plans.json
index fa844560b..5c358f4b5 100644
--- a/tests/fixtures/perf_corpus/golden_plans.json
+++ b/tests/fixtures/perf_corpus/golden_plans.json
@@ -52,6 +52,7 @@
"line_start": 42
},
{
+ "call": "save",
"file_path": "src/app/jobs.py",
"function_name": "drain",
"line_end": 44,
@@ -62,7 +63,7 @@
"api": null,
"evidence_truncated": false,
"intervention_symbol": "src/app/jobs.py::drain",
- "judgment_steps": 4,
+ "judgment_steps": 3,
"mechanical_steps": 0,
"opportunity_id": "perf3_d41b04897cfde2a9eed6",
"paths": [
@@ -75,35 +76,27 @@
"safety": "advisory",
"steps": [
{
- "action": "Add a form of save that takes every key at once",
- "applicability": "judgment",
- "file_path": "src/app/db.py",
- "line": null,
- "order": 1,
- "symbol": "src/app/db.py::save"
- },
- {
- "action": "Collect the keys before the loop and call the batched form once",
+ "action": "Collect the keys before the loop and replace the per-key call with one batched call",
"applicability": "judgment",
"file_path": "src/app/jobs.py",
"line": 40,
- "order": 2,
+ "order": 1,
"symbol": "drain"
},
{
- "action": "Collect the keys before the loop and call the batched form once",
+ "action": "Collect the keys before the loop and replace the per-key call with one batched call",
"applicability": "judgment",
"file_path": "src/app/jobs.py",
"line": 42,
- "order": 3,
+ "order": 2,
"symbol": "drain"
},
{
- "action": "Collect the keys before the loop and call the batched form once",
+ "action": "Collect the keys before the loop and replace the per-key save call with one batched call",
"applicability": "judgment",
"file_path": "src/app/jobs.py",
"line": 44,
- "order": 4,
+ "order": 3,
"symbol": "drain"
}
],
@@ -152,18 +145,21 @@
"plan": {
"affected_locations": [
{
+ "call": "_move",
"file_path": "src/app/migrate.py",
"function_name": "apply",
"line_end": 587,
"line_start": 587
},
{
+ "call": "_fold_one",
"file_path": "src/app/migrate.py",
"function_name": "apply",
"line_end": 590,
"line_start": 590
},
{
+ "call": "_write_alias",
"file_path": "src/app/migrate.py",
"function_name": "apply",
"line_end": 591,
@@ -174,7 +170,7 @@
"api": null,
"evidence_truncated": false,
"intervention_symbol": "src/app/migrate.py::apply",
- "judgment_steps": 4,
+ "judgment_steps": 3,
"mechanical_steps": 0,
"opportunity_id": "perf3_b03c60fe41fa017130c6",
"paths": [
@@ -195,35 +191,27 @@
"safety": "advisory",
"steps": [
{
- "action": "Add a form of the per-key call that takes every key at once",
- "applicability": "judgment",
- "file_path": null,
- "line": null,
- "order": 1,
- "symbol": null
- },
- {
- "action": "Collect the keys before the loop and call the batched form once",
+ "action": "Collect the keys before the loop and replace the per-key _move call with one batched call",
"applicability": "judgment",
"file_path": "src/app/migrate.py",
"line": 587,
- "order": 2,
+ "order": 1,
"symbol": "apply"
},
{
- "action": "Collect the keys before the loop and call the batched form once",
+ "action": "Collect the keys before the loop and replace the per-key _fold_one call with one batched call",
"applicability": "judgment",
"file_path": "src/app/migrate.py",
"line": 590,
- "order": 3,
+ "order": 2,
"symbol": "apply"
},
{
- "action": "Collect the keys before the loop and call the batched form once",
+ "action": "Collect the keys before the loop and replace the per-key _write_alias call with one batched call",
"applicability": "judgment",
"file_path": "src/app/migrate.py",
"line": 591,
- "order": 4,
+ "order": 3,
"symbol": "apply"
}
],
@@ -272,12 +260,14 @@
"plan": {
"affected_locations": [
{
+ "call": "_hit",
"file_path": "src/app/entry.py",
"function_name": "a",
"line_end": 252,
"line_start": 250
},
{
+ "call": "_hit",
"file_path": "src/app/entry.py",
"function_name": "b",
"line_end": 257,
@@ -315,7 +305,7 @@
"symbol": "src/app/entry.py::_hit"
},
{
- "action": "Collect the keys before the loop and call the batched form once",
+ "action": "Collect the keys before the loop and replace the per-key _hit call with one batched call",
"applicability": "judgment",
"file_path": "src/app/entry.py",
"line": 250,
@@ -323,7 +313,7 @@
"symbol": "a"
},
{
- "action": "Collect the keys before the loop and call the batched form once",
+ "action": "Collect the keys before the loop and replace the per-key _hit call with one batched call",
"applicability": "judgment",
"file_path": "src/app/entry.py",
"line": 255,
@@ -376,12 +366,14 @@
"plan": {
"affected_locations": [
{
+ "call": "fetch_page",
"file_path": "src/app/api.py",
"function_name": "list_users",
"line_end": 32,
"line_start": 30
},
{
+ "call": "fetch_page",
"file_path": "src/app/api.py",
"function_name": "list_orders",
"line_end": 33,
@@ -419,7 +411,7 @@
"symbol": "src/app/repo.py::fetch_page"
},
{
- "action": "Collect the keys before the loop and call the batched form once",
+ "action": "Collect the keys before the loop and replace the per-key fetch_page call with one batched call",
"applicability": "judgment",
"file_path": "src/app/api.py",
"line": 30,
@@ -427,7 +419,7 @@
"symbol": "list_users"
},
{
- "action": "Collect the keys before the loop and call the batched form once",
+ "action": "Collect the keys before the loop and replace the per-key fetch_page call with one batched call",
"applicability": "judgment",
"file_path": "src/app/api.py",
"line": 31,
@@ -481,12 +473,14 @@
"plan": {
"affected_locations": [
{
+ "call": "_write",
"file_path": "src/app/reindex.py",
"function_name": "one",
"line_end": 212,
"line_start": 210
},
{
+ "call": "_write",
"file_path": "src/app/reindex.py",
"function_name": "many",
"line_end": 217,
@@ -524,7 +518,7 @@
"symbol": "src/app/reindex.py::_write"
},
{
- "action": "Collect the keys before the loop and call the batched form once",
+ "action": "Collect the keys before the loop and replace the per-key _write call with one batched call",
"applicability": "judgment",
"file_path": "src/app/reindex.py",
"line": 210,
@@ -532,7 +526,7 @@
"symbol": "one"
},
{
- "action": "Collect the keys before the loop and call the batched form once",
+ "action": "Collect the keys before the loop and replace the per-key _write call with one batched call",
"applicability": "judgment",
"file_path": "src/app/reindex.py",
"line": 215,
@@ -656,12 +650,14 @@
"plan": {
"affected_locations": [
{
+ "call": "_load",
"file_path": "src/app/sync.py",
"function_name": "sync_one",
"line_end": 72,
"line_start": 70
},
{
+ "call": "_load",
"file_path": "src/app/sync.py",
"function_name": "sync_all",
"line_end": 73,
@@ -699,7 +695,7 @@
"symbol": "src/app/sync.py::_load"
},
{
- "action": "Collect the keys before the loop and call the batched form once",
+ "action": "Collect the keys before the loop and replace the per-key _load call with one batched call",
"applicability": "judgment",
"file_path": "src/app/sync.py",
"line": 70,
@@ -707,7 +703,7 @@
"symbol": "sync_one"
},
{
- "action": "Collect the keys before the loop and call the batched form once",
+ "action": "Collect the keys before the loop and replace the per-key _load call with one batched call",
"applicability": "judgment",
"file_path": "src/app/sync.py",
"line": 71,
@@ -760,6 +756,7 @@
"plan": {
"affected_locations": [
{
+ "call": "write",
"file_path": "src/app/cache.py",
"function_name": "refresh",
"line_end": 112,
@@ -854,7 +851,7 @@
"safety": "proven",
"steps": [
{
- "action": "Run the independent awaits concurrently, each still inside self._sem",
+ "action": "Run the independent calls concurrently, each still inside self._sem",
"applicability": "mechanical",
"file_path": "src/app/fetch.py",
"line": 40,
@@ -996,7 +993,7 @@
"safety": "advisory",
"steps": [
{
- "action": "Run the independent awaits concurrently, with a bound on how many at once",
+ "action": "Run the independent calls concurrently, with a bound on how many at once",
"applicability": "judgment",
"file_path": "src/app/fanout.py",
"line": 90,
@@ -1049,6 +1046,7 @@
"plan": {
"affected_locations": [
{
+ "call": "get_session",
"file_path": "src/app/job.py",
"function_name": "run",
"line_end": 202,
@@ -1080,7 +1078,7 @@
"symbol": "src/app/db.py::get_session"
},
{
- "action": "Collect the keys before the loop and call the batched form once",
+ "action": "Collect the keys before the loop and replace the per-key get_session call with one batched call",
"applicability": "judgment",
"file_path": "src/app/job.py",
"line": 200,
@@ -1163,7 +1161,7 @@
"symbol": "src/app/db.py::only_sink"
},
{
- "action": "Collect the keys before the loop and call the batched form once",
+ "action": "Collect the keys before the loop and replace the per-key call with one batched call",
"applicability": "judgment",
"file_path": "src/app/lone.py",
"line": 9,
@@ -1340,6 +1338,7 @@
"plan": {
"affected_locations": [
{
+ "call": "get_session",
"file_path": "src/app/orders.py",
"function_name": "handle",
"line_end": 12,
@@ -1371,7 +1370,7 @@
"symbol": "src/app/db.py::get_session"
},
{
- "action": "Collect the keys before the loop and call the batched form once",
+ "action": "Collect the keys before the loop and replace the per-key get_session call with one batched call",
"applicability": "judgment",
"file_path": "src/app/orders.py",
"line": 10,
@@ -1424,6 +1423,7 @@
"plan": {
"affected_locations": [
{
+ "call": "get_session",
"file_path": "src/app/reports.py",
"function_name": "handle",
"line_end": 14,
@@ -1455,7 +1455,7 @@
"symbol": "src/app/db.py::get_session"
},
{
- "action": "Collect the keys before the loop and call the batched form once",
+ "action": "Collect the keys before the loop and replace the per-key get_session call with one batched call",
"applicability": "judgment",
"file_path": "src/app/reports.py",
"line": 12,
@@ -1508,6 +1508,7 @@
"plan": {
"affected_locations": [
{
+ "call": "get_session",
"file_path": "src/app/billing.py",
"function_name": "handle",
"line_end": 13,
@@ -1539,7 +1540,7 @@
"symbol": "src/app/db.py::get_session"
},
{
- "action": "Collect the keys before the loop and call the batched form once",
+ "action": "Collect the keys before the loop and replace the per-key get_session call with one batched call",
"applicability": "judgment",
"file_path": "src/app/billing.py",
"line": 11,
@@ -1602,7 +1603,7 @@
"api": null,
"evidence_truncated": false,
"intervention_symbol": "src/app/retry.py::post_with_retry",
- "judgment_steps": 2,
+ "judgment_steps": 1,
"mechanical_steps": 0,
"opportunity_id": "perf3_bd36a05060a6d02a1245",
"paths": [],
@@ -1610,19 +1611,11 @@
"safety": "advisory",
"steps": [
{
- "action": "Add a form of the per-key call that takes every key at once",
- "applicability": "judgment",
- "file_path": null,
- "line": null,
- "order": 1,
- "symbol": null
- },
- {
- "action": "Collect the keys before the loop and call the batched form once",
+ "action": "Collect the keys before the loop and replace the per-key call with one batched call",
"applicability": "judgment",
"file_path": "src/app/retry.py",
"line": 50,
- "order": 2,
+ "order": 1,
"symbol": "post_with_retry"
}
],
@@ -1671,6 +1664,7 @@
"plan": {
"affected_locations": [
{
+ "call": "get_session",
"file_path": "scripts/backfill.py",
"function_name": "run",
"line_end": 202,
@@ -1702,7 +1696,7 @@
"symbol": "src/app/db.py::get_session"
},
{
- "action": "Collect the keys before the loop and call the batched form once",
+ "action": "Collect the keys before the loop and replace the per-key get_session call with one batched call",
"applicability": "judgment",
"file_path": "scripts/backfill.py",
"line": 200,
@@ -1755,6 +1749,7 @@
"plan": {
"affected_locations": [
{
+ "call": "reliable_sink",
"file_path": "src/app/edge.py",
"function_name": "run",
"line_end": 232,
@@ -1786,7 +1781,7 @@
"symbol": "src/app/db.py::reliable_sink"
},
{
- "action": "Collect the keys before the loop and call the batched form once",
+ "action": "Collect the keys before the loop and replace the per-key reliable_sink call with one batched call",
"applicability": "judgment",
"file_path": "src/app/edge.py",
"line": 230,
@@ -1839,6 +1834,7 @@
"plan": {
"affected_locations": [
{
+ "call": "guessed_sink",
"file_path": "src/app/guess.py",
"function_name": "run",
"line_end": 242,
@@ -1870,7 +1866,7 @@
"symbol": "src/app/db.py::guessed_sink"
},
{
- "action": "Collect the keys before the loop and call the batched form once",
+ "action": "Collect the keys before the loop and replace the per-key guessed_sink call with one batched call",
"applicability": "judgment",
"file_path": "src/app/guess.py",
"line": 240,
diff --git a/tests/unit/health/test_perf_plan_steps.py b/tests/unit/health/test_perf_plan_steps.py
new file mode 100644
index 000000000..61eb093a5
--- /dev/null
+++ b/tests/unit/health/test_perf_plan_steps.py
@@ -0,0 +1,140 @@
+"""Performance plan steps: one per call site, each naming the call it batches.
+
+Two markers observing one call used to become two identical steps, and the
+intro step added a bulk form to whichever member's sink happened to resolve,
+even when the loop's own N+1 was a different call.
+"""
+
+from __future__ import annotations
+
+from repowise.core.analysis.health.complexity import walk_file
+from repowise.core.analysis.health.perf.opportunities import build_performance_opportunities
+from repowise.core.analysis.health.refactoring.performance_fix import (
+ performance_fix_suggestions,
+)
+from tests.unit.health.test_perf_opportunities import _finding
+
+_EVOLUTION = "core/decisions/evolution.py"
+
+
+def _steps(rows) -> list[dict]:
+ opportunities = build_performance_opportunities(rows)
+ assert len(opportunities) == 1
+ return performance_fix_suggestions(opportunities)[0].plan["steps"]
+
+
+def _hits(source: str):
+ return walk_file("t.py", "python", source.encode()).perf_hits
+
+
+def test_the_walker_stores_the_call_each_iteration_makes() -> None:
+ src = (
+ "from sqlalchemy import select\n"
+ "async def sync(session, ids, groups):\n"
+ " for group in groups:\n"
+ " for key in ids:\n"
+ " await session.get(Record, key)\n"
+ " conn.execute(select(key)).fetchone()\n"
+ )
+ calls = {(h.kind, h.line): h.loop_facts().get("sink_call") for h in _hits(src)}
+ assert calls[("io_in_loop", 5)] == "session.get"
+ assert calls[("nested_loop_with_io", 5)] == "session.get"
+ assert calls[("io_in_loop", 6)] == "conn.execute().fetchone"
+
+
+def test_one_site_seen_twice_and_a_helper_sink_make_one_step_per_site() -> None:
+ """The evolution.py shape: two markers on ``session.get`` at one line, and a
+ helper call lower in the same loop whose path ends in ``upsert_decision_edge``."""
+ rows = [
+ _finding(_EVOLUTION, 471, loop_line=467, sink_call="session.get"),
+ _finding(
+ _EVOLUTION, 471, marker="nested_loop_with_io", loop_line=467, sink_call="session.get"
+ ),
+ _finding(
+ _EVOLUTION,
+ 522,
+ loop_line=467,
+ call_path=(f"{_EVOLUTION}::run", "core/graph.py::upsert_decision_edge"),
+ ),
+ ]
+ steps = _steps(rows)
+
+ assert [(s["line"], s["loop_line"]) for s in steps] == [(471, 467), (522, 467)]
+ assert not any("upsert_decision_edge" in str(s["symbol"]) for s in steps)
+ assert "session.get" in steps[0]["action"]
+ assert "upsert_decision_edge" in steps[1]["action"]
+ assert all(s["symbol"] for s in steps)
+
+
+def test_sites_that_reach_one_sink_keep_the_bulk_form_step_and_every_site() -> None:
+ sink = ("shared.py::load", "git.py::head")
+ rows = [
+ _finding("a.py", 10, loop_line=8, call_path=("a.py::run", *sink)),
+ _finding("b.py", 20, loop_line=18, call_path=("b.py::run", *sink)),
+ _finding("b.py", 25, loop_line=18, call_path=("b.py::run", *sink)),
+ ]
+ steps = _steps(rows)
+
+ assert steps[0]["symbol"] == "shared.py::load"
+ assert steps[0]["line"] is None
+ assert [(s["file_path"], s["line"]) for s in steps[1:]] == [
+ ("a.py", 10),
+ ("b.py", 20),
+ ("b.py", 25),
+ ]
+ assert all("load call" in s["action"] for s in steps[1:])
+
+
+def test_direct_calls_with_no_sink_get_no_unnamed_bulk_form_step() -> None:
+ rows = [
+ _finding("jobs.py", 12, loop_line=10, sink_call="session.execute"),
+ _finding("jobs.py", 15, loop_line=10, sink_call="session.commit"),
+ ]
+ steps = _steps(rows)
+
+ assert [s["line"] for s in steps] == [12, 15]
+ assert all(s["symbol"] == "run" for s in steps)
+ assert "session.execute" in steps[0]["action"]
+ assert "session.commit" in steps[1]["action"]
+
+
+def test_a_sink_past_the_evidence_cap_still_withholds_the_bulk_form_step() -> None:
+ rows = [
+ _finding("a.py", line, loop_line=5, call_path=("a.py::run", "db.py::query"))
+ for line in range(10, 90, 10)
+ ]
+ rows.append(_finding("a.py", 95, loop_line=5, call_path=("a.py::run", "db.py::other")))
+ opportunities = build_performance_opportunities(rows)
+ assert len(opportunities) == 1 and opportunities[0].evidence_truncated
+ steps = performance_fix_suggestions(opportunities)[0].plan["steps"]
+
+ assert all(s["line"] for s in steps)
+
+
+def test_a_callee_holding_a_string_literal_gives_its_method() -> None:
+ src = (
+ "from sqlalchemy import select\n"
+ "def sync(session, ids):\n"
+ " for key in ids:\n"
+ " getattr(session, 'get(')(key).execute(select(key))\n"
+ )
+ calls = {h.loop_facts().get("sink_call") for h in _hits(src) if h.kind == "io_in_loop"}
+ assert calls == {"execute"}
+
+
+def test_java_and_kotlin_name_the_receiver() -> None:
+ java = (
+ "class A { void f(java.sql.Connection conn, java.util.List<String> ids) throws Exception {\n"
+ " for (String id : ids) { conn.prepareStatement(id).executeQuery(); }\n"
+ "} }\n"
+ )
+ kotlin = (
+ "import java.sql.Connection\n"
+ "fun f(conn: Connection, ids: List<String>) {\n"
+ " for (id in ids) { conn.prepareStatement(id).executeQuery() }\n"
+ "}\n"
+ )
+ for path, language, src in (("A.java", "java", java), ("A.kt", "kotlin", kotlin)):
+ hits = walk_file(path, language, src.encode()).perf_hits
+ calls = {h.loop_facts().get("sink_call") for h in hits if h.kind == "io_in_loop"}
+ assert calls == {"conn.prepareStatement().executeQuery"}, language