Skip to content

Impact tests

impact_tests

Symbol-aware test discovery for impact analysis, direct and transitive.

ast_impact reports two families of test files for an edited symbol:

  • direct: test files that reference the symbol. A function matches on its name as a whole identifier. A method also needs evidence that the test can reach the owning class: the class name (or one of its subclasses) appears in the file, or the file imports the defining module. Dunder methods (__init__, __call__...) are invoked implicitly, so the class reference alone is enough for them.
  • transitive (opt-in): the direct tests of the symbol and of every in-package caller reached by walking the call graph upward, level by level, up to a bounded depth. The walk is cycle-safe (each symbol is visited once) and capped at :data:MAX_TRANSITIVE_SYMBOLS visited symbols. Hitting the cap is reported, never silent.

The test suite is read and tokenised once per analysis (:class:TestCorpus), so every lookup during the walk is a set intersection, not a rescan of the suite.

ClassHierarchy

Name-based class hierarchy of one package.

Bases are matched by bare name, so two homonymous classes in distinct modules share a node. That is a deliberate over-approximation: it can only widen the owner family, never drop a real subclass.

Source code in packages/axm-ast/src/axm_ast/core/impact_tests.py
Python
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
class ClassHierarchy:
    """Name-based class hierarchy of one package.

    Bases are matched by bare name, so two homonymous classes in distinct
    modules share a node. That is a deliberate over-approximation: it can
    only widen the owner family, never drop a real subclass.
    """

    def __init__(self, pkg: PackageInfo) -> None:
        """Index every class of *pkg* by name and by base name."""
        self._bases: dict[str, set[str]] = {}
        self._children: dict[str, set[str]] = {}
        for mod in pkg.modules:
            for cls in mod.classes:
                bases = {_bare_base(base) for base in cls.bases}
                self._bases.setdefault(cls.name, set()).update(bases)
                for base in bases:
                    self._children.setdefault(base, set()).add(cls.name)

    @staticmethod
    def _closure(start: str, edges: dict[str, set[str]]) -> frozenset[str]:
        seen = {start}
        stack = [start]
        while stack:
            for nxt in edges.get(stack.pop(), ()):
                if nxt not in seen:
                    seen.add(nxt)
                    stack.append(nxt)
        return frozenset(seen)

    def descendants(self, name: str) -> frozenset[str]:
        """Return *name* and every class inheriting from it."""
        return self._closure(name, self._children)

    def family(self, name: str) -> frozenset[str]:
        """Return *name*, its ancestors and its descendants.

        A ``self.m()`` call in a base class may dispatch to an override, and
        one in a subclass may reach the inherited definition.
        """
        return self._closure(name, self._bases) | self.descendants(name)
__init__(pkg)

Index every class of pkg by name and by base name.

Source code in packages/axm-ast/src/axm_ast/core/impact_tests.py
Python
101
102
103
104
105
106
107
108
109
110
def __init__(self, pkg: PackageInfo) -> None:
    """Index every class of *pkg* by name and by base name."""
    self._bases: dict[str, set[str]] = {}
    self._children: dict[str, set[str]] = {}
    for mod in pkg.modules:
        for cls in mod.classes:
            bases = {_bare_base(base) for base in cls.bases}
            self._bases.setdefault(cls.name, set()).update(bases)
            for base in bases:
                self._children.setdefault(base, set()).add(cls.name)
descendants(name)

Return name and every class inheriting from it.

Source code in packages/axm-ast/src/axm_ast/core/impact_tests.py
Python
123
124
125
def descendants(self, name: str) -> frozenset[str]:
    """Return *name* and every class inheriting from it."""
    return self._closure(name, self._children)
family(name)

Return name, its ancestors and its descendants.

A self.m() call in a base class may dispatch to an override, and one in a subclass may reach the inherited definition.

Source code in packages/axm-ast/src/axm_ast/core/impact_tests.py
Python
127
128
129
130
131
132
133
def family(self, name: str) -> frozenset[str]:
    """Return *name*, its ancestors and its descendants.

    A ``self.m()`` call in a base class may dispatch to an override, and
    one in a subclass may reach the inherited definition.
    """
    return self._closure(name, self._bases) | self.descendants(name)

DiscoveryContext dataclass

Everything a transitive walk needs, computed once per analysis.

Attributes:

Name Type Description
pkg PackageInfo

Package whose call graph is walked.
corpus TestCorpus

Test corpus the direct matches are taken from.
reaches_module Callable[[Path, str], bool]

(test_file, module) -> bool import predicate.
keep_function_callers Callable[[SymbolTarget, list[CallSite]], list[CallSite]]

Provenance filter applied to the call sites of a plain function (drops callers proven to import a homonym).
Source code in packages/axm-ast/src/axm_ast/core/impact_tests.py
Python
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
@dataclass
class DiscoveryContext:
    """Everything a transitive walk needs, computed once per analysis.

    Attributes:
        pkg: Package whose call graph is walked.
        corpus: Test corpus the direct matches are taken from.
        reaches_module: ``(test_file, module) -> bool`` import predicate.
        keep_function_callers: Provenance filter applied to the call sites
            of a plain function (drops callers proven to import a homonym).
    """

    pkg: PackageInfo
    corpus: TestCorpus
    reaches_module: Callable[[Path, str], bool]
    keep_function_callers: Callable[[SymbolTarget, list[CallSite]], list[CallSite]]
    hierarchy: ClassHierarchy = field(init=False)
    _modules: dict[str, ModuleInfo] = field(init=False)
    _sources: dict[str, frozenset[str]] = field(init=False, default_factory=dict)
    _direct: dict[SymbolTarget, list[Path]] = field(init=False, default_factory=dict)

    def __post_init__(self) -> None:
        """Index the package once."""
        self.hierarchy = ClassHierarchy(self.pkg)
        self._modules = {
            module_dotted_name(mod.path, self.pkg.root): mod for mod in self.pkg.modules
        }

    def direct_tests(self, target: SymbolTarget) -> list[Path]:
        """Direct test files of *target* (memoised)."""
        if target not in self._direct:
            self._direct[target] = self.corpus.matching(
                target, self.hierarchy, self.reaches_module
            )
        return self._direct[target]

    def module_tokens(self, module: str) -> frozenset[str]:
        """Identifiers of a package module's source (memoised, empty on error)."""
        if module not in self._sources:
            mod = self._modules.get(module)
            try:
                text = mod.path.read_text(encoding="utf-8") if mod else ""
            except OSError:
                text = ""
            self._sources[module] = frozenset(_IDENTIFIER.findall(text))
        return self._sources[module]

    def enclosing(self, site: CallSite) -> SymbolTarget | None:
        """Top-level function or method whose body holds *site*."""
        mod = self._modules.get(site.module)
        if mod is None:
            return None
        for fn in mod.functions:
            if fn.line_start <= site.line <= fn.line_end:
                return SymbolTarget(site.module, fn.name)
        for cls in mod.classes:
            for method in cls.methods:
                if method.line_start <= site.line <= method.line_end:
                    return SymbolTarget(site.module, f"{cls.name}.{method.name}")
        return None

    def enclosing_class(self, site: CallSite) -> ClassInfo | None:
        """Top-level class whose body holds *site*."""
        mod = self._modules.get(site.module)
        if mod is None:
            return None
        return next(
            (c for c in mod.classes if c.line_start <= site.line <= c.line_end),
            None,
        )
__post_init__()

Index the package once.

Source code in packages/axm-ast/src/axm_ast/core/impact_tests.py
Python
221
222
223
224
225
226
def __post_init__(self) -> None:
    """Index the package once."""
    self.hierarchy = ClassHierarchy(self.pkg)
    self._modules = {
        module_dotted_name(mod.path, self.pkg.root): mod for mod in self.pkg.modules
    }
direct_tests(target)

Direct test files of target (memoised).

Source code in packages/axm-ast/src/axm_ast/core/impact_tests.py
Python
228
229
230
231
232
233
234
def direct_tests(self, target: SymbolTarget) -> list[Path]:
    """Direct test files of *target* (memoised)."""
    if target not in self._direct:
        self._direct[target] = self.corpus.matching(
            target, self.hierarchy, self.reaches_module
        )
    return self._direct[target]
enclosing(site)

Top-level function or method whose body holds site.

Source code in packages/axm-ast/src/axm_ast/core/impact_tests.py
Python
247
248
249
250
251
252
253
254
255
256
257
258
259
def enclosing(self, site: CallSite) -> SymbolTarget | None:
    """Top-level function or method whose body holds *site*."""
    mod = self._modules.get(site.module)
    if mod is None:
        return None
    for fn in mod.functions:
        if fn.line_start <= site.line <= fn.line_end:
            return SymbolTarget(site.module, fn.name)
    for cls in mod.classes:
        for method in cls.methods:
            if method.line_start <= site.line <= method.line_end:
                return SymbolTarget(site.module, f"{cls.name}.{method.name}")
    return None
enclosing_class(site)

Top-level class whose body holds site.

Source code in packages/axm-ast/src/axm_ast/core/impact_tests.py
Python
261
262
263
264
265
266
267
268
269
def enclosing_class(self, site: CallSite) -> ClassInfo | None:
    """Top-level class whose body holds *site*."""
    mod = self._modules.get(site.module)
    if mod is None:
        return None
    return next(
        (c for c in mod.classes if c.line_start <= site.line <= c.line_end),
        None,
    )
module_tokens(module)

Identifiers of a package module's source (memoised, empty on error).

Source code in packages/axm-ast/src/axm_ast/core/impact_tests.py
Python
236
237
238
239
240
241
242
243
244
245
def module_tokens(self, module: str) -> frozenset[str]:
    """Identifiers of a package module's source (memoised, empty on error)."""
    if module not in self._sources:
        mod = self._modules.get(module)
        try:
            text = mod.path.read_text(encoding="utf-8") if mod else ""
        except OSError:
            text = ""
        self._sources[module] = frozenset(_IDENTIFIER.findall(text))
    return self._sources[module]

SymbolTarget dataclass

A resolved symbol: package-relative module plus qualified name.

qualname is "func" for a top-level function or class and "Class.method" for a method.

Source code in packages/axm-ast/src/axm_ast/core/impact_tests.py
Python
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
@dataclass(frozen=True)
class SymbolTarget:
    """A resolved symbol: package-relative module plus qualified name.

    ``qualname`` is ``"func"`` for a top-level function or class and
    ``"Class.method"`` for a method.
    """

    module: str
    qualname: str

    @property
    def name(self) -> str:
        """Bare name as it appears at call sites."""
        return self.qualname.rsplit(".", 1)[-1]

    @property
    def owner(self) -> str | None:
        """Owning class name for a method, ``None`` otherwise."""
        head, sep, _ = self.qualname.rpartition(".")
        return head.rsplit(".", 1)[-1] if sep else None

    @property
    def key(self) -> str:
        """Fully qualified identity, ``module.qualname``."""
        return f"{self.module}.{self.qualname}" if self.module else self.qualname
key property

Fully qualified identity, module.qualname.

name property

Bare name as it appears at call sites.

owner property

Owning class name for a method, None otherwise.

TestCorpus

The test files of a project, read and tokenised once.

Matching is by whole identifier: greet matches greet( or mod.greet but not greeting.

Source code in packages/axm-ast/src/axm_ast/core/impact_tests.py
Python
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
class TestCorpus:
    """The test files of a project, read and tokenised once.

    Matching is by whole identifier: ``greet`` matches ``greet(`` or
    ``mod.greet`` but not ``greeting``.
    """

    __test__ = False  # not a pytest test class despite the name

    def __init__(self, project_root: Path) -> None:
        """Read every ``test_*.py`` of the configured suites under *project_root*."""
        self.root = project_root
        self.files: list[Path] = sorted(
            test_file
            for tests_dir in resolve_suite_dirs(project_root)
            for test_file in tests_dir.rglob("test_*.py")
        )
        self._index: dict[str, set[int]] = {}
        for position, test_file in enumerate(self.files):
            try:
                content = test_file.read_text(encoding="utf-8")
            except OSError:
                continue
            for token in set(_IDENTIFIER.findall(content)):
                self._index.setdefault(token, set()).add(position)

    def _referencing(self, names: Iterable[str]) -> set[int]:
        found: set[int] = set()
        for name in names:
            found |= self._index.get(name, set())
        return found

    def referencing(self, name: str) -> list[Path]:
        """Return the test files mentioning *name* as an identifier."""
        return [self.files[i] for i in sorted(self._referencing([name]))]

    def matching(
        self,
        target: SymbolTarget,
        hierarchy: ClassHierarchy,
        reaches_module: Callable[[Path, str], bool],
    ) -> list[Path]:
        """Return the test files directly referencing *target*.

        Args:
            target: Resolved symbol.
            hierarchy: Class hierarchy used to accept subclass references.
            reaches_module: ``(test_file, module) -> bool`` predicate telling
                whether a test imports the defining module (fail-open).
        """
        owner = target.owner
        if owner is None:
            return self.referencing(target.name)
        owner_refs = self._referencing(hierarchy.descendants(owner))
        if _is_dunder(target.name):
            return [self.files[i] for i in sorted(owner_refs)]
        name_refs = self._referencing([target.name])
        accepted = name_refs & owner_refs
        for position in sorted(name_refs - owner_refs):
            if reaches_module(self.files[position], target.module):
                accepted.add(position)
        return [self.files[i] for i in sorted(accepted)]
__init__(project_root)

Read every test_*.py of the configured suites under project_root.

Source code in packages/axm-ast/src/axm_ast/core/impact_tests.py
Python
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
def __init__(self, project_root: Path) -> None:
    """Read every ``test_*.py`` of the configured suites under *project_root*."""
    self.root = project_root
    self.files: list[Path] = sorted(
        test_file
        for tests_dir in resolve_suite_dirs(project_root)
        for test_file in tests_dir.rglob("test_*.py")
    )
    self._index: dict[str, set[int]] = {}
    for position, test_file in enumerate(self.files):
        try:
            content = test_file.read_text(encoding="utf-8")
        except OSError:
            continue
        for token in set(_IDENTIFIER.findall(content)):
            self._index.setdefault(token, set()).add(position)
matching(target, hierarchy, reaches_module)

Return the test files directly referencing target.

Parameters:

Name Type Description Default
target SymbolTarget

Resolved symbol.

 required
hierarchy ClassHierarchy

Class hierarchy used to accept subclass references.

 required
reaches_module Callable[[Path, str], bool]

(test_file, module) -> bool predicate telling whether a test imports the defining module (fail-open).

 required
Source code in packages/axm-ast/src/axm_ast/core/impact_tests.py
Python
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
def matching(
    self,
    target: SymbolTarget,
    hierarchy: ClassHierarchy,
    reaches_module: Callable[[Path, str], bool],
) -> list[Path]:
    """Return the test files directly referencing *target*.

    Args:
        target: Resolved symbol.
        hierarchy: Class hierarchy used to accept subclass references.
        reaches_module: ``(test_file, module) -> bool`` predicate telling
            whether a test imports the defining module (fail-open).
    """
    owner = target.owner
    if owner is None:
        return self.referencing(target.name)
    owner_refs = self._referencing(hierarchy.descendants(owner))
    if _is_dunder(target.name):
        return [self.files[i] for i in sorted(owner_refs)]
    name_refs = self._referencing([target.name])
    accepted = name_refs & owner_refs
    for position in sorted(name_refs - owner_refs):
        if reaches_module(self.files[position], target.module):
            accepted.add(position)
    return [self.files[i] for i in sorted(accepted)]
referencing(name)

Return the test files mentioning name as an identifier.

Source code in packages/axm-ast/src/axm_ast/core/impact_tests.py
Python
168
169
170
def referencing(self, name: str) -> list[Path]:
    """Return the test files mentioning *name* as an identifier."""
    return [self.files[i] for i in sorted(self._referencing([name]))]

TransitiveTests dataclass

Outcome of a transitive walk.

Attributes:

Name Type Description
test_files list[Path]

Direct tests of the target and of every walked caller.
callers list[str]

Walked caller symbols (module.qualname), target excluded.
truncated bool

True when :data:MAX_TRANSITIVE_SYMBOLS stopped the walk before depth was exhausted.
Source code in packages/axm-ast/src/axm_ast/core/impact_tests.py
Python
343
344
345
346
347
348
349
350
351
352
353
354
355
356
@dataclass(frozen=True)
class TransitiveTests:
    """Outcome of a transitive walk.

    Attributes:
        test_files: Direct tests of the target and of every walked caller.
        callers: Walked caller symbols (``module.qualname``), target excluded.
        truncated: ``True`` when :data:`MAX_TRANSITIVE_SYMBOLS` stopped the
            walk before *depth* was exhausted.
    """

    test_files: list[Path]
    callers: list[str]
    truncated: bool

find_transitive_tests(ctx, target, depth, *, max_symbols=MAX_TRANSITIVE_SYMBOLS)

Walk the callers of target upward and collect their direct tests.

Breadth-first, one level per depth unit, each symbol visited once (cycle-safe). depth=0 returns the direct tests only.

Parameters:

Name Type Description Default
ctx DiscoveryContext

Shared discovery context.

 required
target SymbolTarget

Symbol whose change is analysed.

 required
depth int

Number of caller levels to climb (0 to :data:MAX_TRANSITIVE_DEPTH).

 required
max_symbols int

Visited-symbol budget, target included.

 MAX_TRANSITIVE_SYMBOLS

Raises:

Type Description
ValueError

If depth is outside [0, MAX_TRANSITIVE_DEPTH].
Source code in packages/axm-ast/src/axm_ast/core/impact_tests.py
Python
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
def find_transitive_tests(
    ctx: DiscoveryContext,
    target: SymbolTarget,
    depth: int,
    *,
    max_symbols: int = MAX_TRANSITIVE_SYMBOLS,
) -> TransitiveTests:
    """Walk the callers of *target* upward and collect their direct tests.

    Breadth-first, one level per depth unit, each symbol visited once
    (cycle-safe). ``depth=0`` returns the direct tests only.

    Args:
        ctx: Shared discovery context.
        target: Symbol whose change is analysed.
        depth: Number of caller levels to climb (``0`` to
            :data:`MAX_TRANSITIVE_DEPTH`).
        max_symbols: Visited-symbol budget, target included.

    Raises:
        ValueError: If *depth* is outside ``[0, MAX_TRANSITIVE_DEPTH]``.
    """
    validate_transitive_depth(depth)
    walk = _Walk(seen={target}, tests=set(ctx.direct_tests(target)))
    frontier = [target]
    for _ in range(depth):
        frontier = _climb(ctx, walk, frontier, max_symbols)
        if walk.truncated or not frontier:
            break
    return TransitiveTests(
        test_files=sorted(walk.tests),
        callers=sorted(t.key for t in walk.seen if t != target),
        truncated=walk.truncated,
    )

validate_transitive_depth(depth)

Reject a transitive depth outside [0, MAX_TRANSITIVE_DEPTH].

Raises:

Type Description
ValueError

If depth is out of range (or not an int).
Source code in packages/axm-ast/src/axm_ast/core/impact_tests.py
Python
327
328
329
330
331
332
333
334
335
336
337
338
339
340
def validate_transitive_depth(depth: int) -> None:
    """Reject a transitive depth outside ``[0, MAX_TRANSITIVE_DEPTH]``.

    Raises:
        ValueError: If *depth* is out of range (or not an ``int``).
    """
    if isinstance(depth, bool) or not isinstance(depth, int):
        msg = f"transitive_depth must be an integer, got {depth!r}"
        raise ValueError(msg)
    if not 0 <= depth <= MAX_TRANSITIVE_DEPTH:
        msg = (
            f"transitive_depth must be within [0, {MAX_TRANSITIVE_DEPTH}], got {depth}"
        )
        raise ValueError(msg)