An ambiguous extension (`.inc`, `.h`, TS-family, ...) is hard-mapped to one
global default; when that is wrong the file does not error — it parses as the
wrong language and yields a few incidental nodes, so the graph looks populated
while the real symbols are missing. A project can now declare what an extension
means via `language.<ext>=<language|.ext>` lines in `.graphifyrc` (which already
holds `viz_node_limit`). The declaration flows through file classification,
extractor selection, grammar/resolution, imports, watch, and the AST cache
(salt + manifest language field + incremental invalidation), is thread-local
per scan root, and is forwarded to the worker pool. Default detection is
byte-for-byte unchanged when no declaration is present.
Landed from PR #3789 (squashed), which carried @abhay-codes07's original #3075
forward with reviewed follow-ups. Authorship credited to both contributors.
Co-Authored-By: Andre Brait <andrebrait@gmail.com>
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The read/search nudge was re-injected on every qualifying Read/Glob/Grep/Bash
call with no dedup or session cap (issue #3435 measured ~8,348 injections,
~651k tokens, for 339 graphify calls), and an agent that just ran
`graphify query` was still nudged on its very next read. `_nudge_allowed`
now gates all three soft-nudge sites with a per-session + per-agent counter
(first nudge always fires; later ones suppressed while the query stamp is
fresh; hard cap via GRAPHIFY_HOOK_NUDGE_CAP, default 5), with 24h GC and
fail-open preserved. Subagents get independent budgets. The strict deny path
is unchanged.
Landed from PR #3443 (squashed). The README strict-mode note is omitted here
because that prose moved to docs.graphify.com in the README trim.
Co-Authored-By: Andre Brait <andrebrait@gmail.com>
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
graph.json was pretty-printed (indent=2), so every PR that commits a
rebuilt graph reports tens of thousands of changed lines. GitHub's PR
line counts ignore .gitattributes (-diff, binary), so only the file's
content can fix it: one line of compact JSON is +1 -1.
Every writer of a project graph.json now emits separators=(",", ":")
with no indent: to_json, the `update` raw/no-cluster candidate text,
prune, merge-driver, merge-graphs, `extract --no-cluster` and the global
graph. Readers are unaffected.
`update` compares parsed data, not bytes, so an existing indented graph
would have been "unchanged" forever. It is now rewritten compact once,
then left alone.
Co-Authored-By: Claude <noreply@anthropic.com>
(cherry picked from commit 4c31235c37)
Give each binding of a use tree the import edge that v8 produces for
its equivalent plain use declaration. Preserve aliases, path comments,
separator whitespace, nested prefixes and self/glob binding behavior.
Keep the old group edge as call evidence until the call pass finishes;
exclude per-binding edges from that pass, then remove private markers.
Plain imports and all call edges retain their current-v8 behavior.
AST cache schema 6 -> 7 rejects results written before expansion.
Replace the earlier broad resolver in #4135 with this syntax-only fix.
The resolver is preserved on local branch wip/rust-use-resolver.
On current v8, all 30 new tests and the full 6642-test suite pass.
Fresh ripgrep builds have 864 import edges instead of 247, with every
call unchanged. Existing name-based import rewiring can still produce
wrong matches, just as it does for equivalent plain declarations.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Codex <noreply@openai.com>
(cherry picked from commit 2669ee4f11)
A committed graphify-out/graph.json is rewritten whole on every build, so
its pull-request diffs are large by design. Assistants reviewing such PRs
flagged them as churn and proposed skipping the rebuild or its hooks. The
Claude, AGENTS.md and Gemini blocks now say the diff is expected.
(cherry picked from commit e6cf17c171)
Persist validated Codex extraction responses as chunk files so the shared B3 merge preserves semantic nodes and edges.
Fail before cache and graph updates when extraction responses are missing or invalid.
Add regression coverage for chunk merging, cached reruns, and failure handling.
Fixes#4311
Co-authored-by: OpenAI Codex <codex@openai.com>
(cherry picked from commit da5850f0ac)
`case class Point(x: Int) derives Show, Eq` (and the same on enums) now
emits a `references` edge with the `derives` context from the type to each
derived typeclass, next to the class-level context bounds it resembles. A
qualified name (`derives cats.Show`) links by its last segment.
Co-Authored-By: Grok Bot <noreply@x.ai>
(cherry picked from commit e6ccd59f20)
global add loaded graph.json with a plain undirected node_link_graph, so
every link whose target precedes its source in the node list was written
to ~/.graphify/global-graph.json reversed. Stamp _src/_tgt with
restore_arc_direction before the load, as the export loader and
load_node_link_graph do since #4088.
The external-node dedup rewired edge endpoints but not the markers, so a
deduplicated repo's markers named a node that no longer exists. Remap
_src/_tgt along with the endpoints.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
(cherry picked from commit 477848390e)
explain listed the 20 most connected neighbors and "... and N more" with
no option to see the rest, and the token-budget notices of query,
get_neighbors and get_community said "raise the token budget" without
the value that would return everything, so a caller retried with a guess.
explain gets --all, and its cut line names the total and the option.
The budget notices now give the smallest budget that returns the whole
output, computed with the same rule that does the cut.
Co-Authored-By: Claude <noreply@anthropic.com>
(cherry picked from commit 769184dc9a)
explain and query exited 0 when no node matched, the same status as an
answer, and path exited 1 for an ambiguous endpoint, a missing one and a
missing graph file alike. A script or an agent could not tell a refusal
from an answer without parsing the text.
explain, path and query now exit 3 when a name is ambiguous (the
candidates are listed) and 4 when no node matches. Messages and output
streams are unchanged; "no path found" between two existing nodes is
still an answer. The codes are documented in --help and the README.
Co-Authored-By: Claude <noreply@anthropic.com>
(cherry picked from commit 13dc2400e1)
Run from a subdirectory of the scan root, a root-relative node path is resolved
against the CWD, so the renamed node's key stops matching its edges' and the
edges stay on the survivor. Fails on v8; fixed in the next commit.
Co-authored-by: Gijs Mos <1526544+gmos@users.noreply.github.com>
(cherry picked from commit f3697420c4)
# Conflicts:
# tests/test_dedup.py
_norm built the dedup key with re.sub(r"[\W_]+", " ", ...). \w does not
match Unicode combining marks (Mn/Mc: vowel signs, viramas), so they were
deleted like punctuation: "हिंदी" became "ह द", and "कि", "की", "के" all
collapsed to "क", merging three distinct nodes in Pass 1.
Treat a character as part of a word when it is alphanumeric or a combining
mark; everything else still collapses to a single space. Output is
unchanged for strings without combining marks (checked against the old
implementation on ~190k random strings).
Known gaps: ZWJ/ZWNJ still act as separators; ids.normalize_id and the
dormant build.deduplicate_by_label keep the old recipe.
(cherry picked from commit ccd81ce818)
- Unwrap `::RetType` (alongside `where`) around a definition's call head, so
`function f(x)::Float64` is no longer dropped.
- Name method definitions via `_def_name`: `Base.show`, `Base.:+`, bare
operators, and functors `(c::Circle)(x)`. The id uses the bare name so all
methods of a generic function share one node; the label keeps the qualifier.
- Emit `calls` edges for broadcast calls (`f.(xs)`) and pipes (`x |> f`, `.|>`).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
(cherry picked from commit cf3648e762)
Some OpenAI-compatible gateways return a usage object whose token-count
fields are null. Reading them verbatim left None in result["input_tokens"] /
result["output_tokens"], which crashed extraction for the whole chunk
(TypeError: '<' not supported between 'NoneType' and 'int' in the ollama
low-token warning; None also propagated through the per-chunk merge).
Coerce null counts to 0 at read time in _call_openai_compat, _call_azure,
_call_claude, and _call_bedrock, matching the existing `or 0` idiom already
used on the claude-cli path.
Fixes#4325
Signed-off-by: Raphael Malikian <rtmalikian@gmail.com>
(cherry picked from commit 640005fdb7)
A set-but-empty GRAPHIFY_OUT (a blank .env / docker-compose / CI entry)
or GRAPHIFY_OUT=. made the purge target <project>/"" == the project
itself, and rmtree deleted the whole project.
- paths: treat a blank GRAPHIFY_OUT as unset; "." / ".." fall back to
graphify-out with a warning. GRAPHIFY_OUT_NAME and out_path() follow.
- install: before any uninstaller runs, refuse --purge (exit 1, nothing
changed) when the target is the project, a parent of it, the home
folder or a filesystem root. This also covers an absolute GRAPHIFY_OUT.
- hooks: the rebuild bodies resolve the output dir via graphify.paths
instead of reading the env var directly.
Fixes#4328
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
(cherry picked from commit 7d6befbb9c)
_collect_js_symbol_resolution_facts walked every parsed tree in full four
times, and each pass kept only a few node types.
The first pass now walks once and keeps the nodes of the six types the
whole-tree passes filter on, in walk order. Later passes iterate that
list. Each fact list is filled by a single pass, so fact order does not
change.
_walk_js_tree now uses a TreeCursor instead of reading .children at every
node. A cursor is bounded by its start node, so the walks that start
inside a tree (function bodies, and the import and export nodes passed to
_js_named_specifiers) stay inside that subtree.
Overlaps #3353, which also indexes JS nodes once and frees each tree
after use.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
(cherry picked from commit 243deb39e7)
_source_key and _resolve_cached key their lru_cache on the path and the
cwd, and call os.getcwd() on every lookup, hit or miss. On a 900-file
repo an extract makes about 215k of those calls. os.getcwd() is a
syscall that costs 30 to 40us on macOS, so building the key cost more
than the resolution the memo saved.
An absolute path resolves the same from any cwd, so it now keys with an
empty cwd. A relative path still keys on the cwd, so a chdir still misses
the cache. Path.is_absolute() applies host rules, which is what local
resolution needs: a driveless path on Windows depends on the drive.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
(cherry picked from commit 2228f80804)
Dart/Flutter (and Python, Go, and other file-module languages) extract a
class or function as a symbol node with a defines edge from its source
file. Without defines in DEFAULT_AFFECTED_RELATIONS the reverse walk
stops at depth 0 because the only incoming edge is invisible.
Measured on a real Flutter project (12 104 nodes, 12 420 edges):
- 4 793 symbols defined via defines edges
- 4 691 of them (97.9 %) had zero incoming edges from any whitelisted
relation — affected returned 'No affected nodes found' for each one
- After adding defines: 100 % of symbols reachable through at minimum
their defining file
Tests: 10 new regression tests in test_affected_defines.py.
All pass post-fix, all fail pre-fix.
(cherry picked from commit c695fdf8eb)
A named generic interface (`interface area; module procedure area_circle,
area_square; end interface`) names a callable that dispatches to its
specific procedures by argument type. The interface block was never turned
into a node, so a call to the generic name dangled — a Fortran function
call_expression only resolves to an in-file node — and the dispatch set
was invisible, even though the specific procedures were extracted. Generic
interfaces (and operator overload sets) are routine in scientific Fortran.
Emit a node for the generic name, link it to the procedures it groups with
`dispatches_to` edges (resolved after the walk, like type-bound
procedures, so a procedure from another module becomes a collapsible
sourceless stub), and the existing call resolution then binds `area(x)`
to it. Unnamed abstract/explicit interfaces keep recursing so their
procedure declarations are still extracted.
(cherry picked from commit ccfb4018fb)
My #4141 fix used a one-off calls/generic_dispatch relation; PR #4142
(@rajatnagda45) implements the same fix with the codebase-standard
dispatches_to relation and earns author-preserving credit. Reverting the
Fortran hunk here so #4142 can be cherry-picked; the Julia fixes from the
same original commit are unaffected.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Landed from PR #4282 (squashed, authorship preserved). Covers the extension-drop
collision (x.py vs x.html); the path-dropped doc-id case in #4281 remains open.
Landed from PR #4251 (squashed, authorship preserved). The PR's README note on
re-inclusion semantics is omitted here since the README's Ignoring files section
now links to docs.graphify.com/reference/configuration.
hook-guard read returned early for .ps1/.psm1/.psd1 targets because the
extension gate (_HOOK_SOURCE_EXTS in graphify/cli.py) did not list them,
even though detect.py classifies those extensions as code and they are
extracted into the graph. In a PowerShell repo the Read/Glob nudge never
fired, only the search one.
FixesGraphify-Labs/graphify#4271.
Signed-off-by: Raphael Malikian <rtmalikian@gmail.com>
Co-Authored-By: Claude <noreply@anthropic.com>
(cherry picked from commit 1ee08e70ba)
A Scala 3 `given ... with` instance's `def`s leaked to file scope, where `show`
(cased to `_show`) collided with the trait TYPE node `Show`, so the given's
call/reference edges landed on the trait. Treat `given_definition` as a
class-like container with a `with_template_body`, and synthesize a stable name
for an anonymous given (`given Show[Int] with`) from its implemented type so its
members are scoped instead of dropped.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Node data is embedded as JSON inside a <script> element, but the serializers
escaped only `</`. An unclosed `<!--` in one label followed by a later `<script`
drove the HTML tokenizer into script-data-double-escaped state, where the real
</script> no longer closed the tag and graph.html rendered blank. Escape every
`<` as < in both exporters/html.py and tree_html.py; labels still render
because `<` round-trips through JSON.parse.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
`defdelegate name(args), to: Mod` declares an invocable function with the same
`call` head shape as `def`, but was missing from the def-family recognizer, so no
node was minted and same-module calls to the delegated name dangled. Add
`defdelegate` to the def-family tuple and to the call-pass skip set.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
- Julia: a `where` clause wraps the call head in a `where_expression` and a
parametric type name sits inside a `parametrized_type_expression`, so the
name-lookup helpers missed both and dropped the method/type entirely. Unwrap
the wrapper nodes in `_func_name_from_signature`, the short-function assignment
branch, and `_type_head_names`. Declared type parameters (`value::T`) no longer
leak as phantom field-type reference nodes (#4126, #4128).
- Fortran: a named generic `interface` had no walk branch, so the generic name
was dropped and a call to it dangled. Emit a node for the generic name and
defer-resolve a `generic_dispatch` edge to each listed specific procedure,
mirroring the type-bound-procedure pattern; unnamed/operator interfaces are
untouched and their bodies still extracted (#4141).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
#4262 reported that a type defined exactly once never bound across modules,
leaving hundreds of per-file stubs. It no longer reproduces on v8: the
reference binds through the re-export chain (and, absent any import, through the
unique-stub rewire). Add regression coverage matching the reporter's node
layout (config/settings/_models.py -> config_settings_models_settings) plus an
import-free inheritance control, so the resolved behavior cannot silently
regress. No production change required.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
An external type the corpus references but never declares (`List<T>`,
`System.Type`, Unity's `VisualElement`) got a separate sourceless placeholder
in every file that mentioned it, with the file's name baked into the id
(`a_cs_list`, `b_cs_list`). Queries returned a dozen identical nodes and the
files sharing the type stayed unconnected.
`_resolve_csharp_type_references` minted/kept a per-file stub because the arbiter
short-circuited on whichever placeholder each edge already pointed at. It now
routes every undeclared same-name reference to a single canonical, name-keyed
stub (mirroring Go's external-type treatment), still shadowing no real or
differently-labelled node. The now-unreferenced per-file placeholders are pruned
by the existing sweep.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
`def read_body(response: httpx.Response)` names an external type, but a private
`class _Response` in a test file absorbed the reference, producing a fabricated
EXTRACTED-confidence edge from production code to test code.
The stub-rewire key (`_node_label_key`) stripped ALL non-alphanumeric
characters, including underscores, so `_Response` and `Response` collapsed to
the same key and the lone local `_Response` was chosen as the unique match.
Underscores are significant identifier characters, so the key now preserves
them (only `()`, `<>`, `.`, and whitespace are dropped); `Foo()` still matches
`Foo`, but `_Response` no longer matches `Response`.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
A non-package directory with a file named like a stdlib module (e.g.
`scripts/logging.py`) wrongly captured a bare `import logging` from a
sibling, binding the import — and member calls like `logging.getLogger()` —
to the local file instead of the stdlib module.
The false binding arose at three independent layers, all now guarded by a
vendored, deterministic `_PYTHON_STDLIB_MODULE_NAMES` frozenset (union of
3.10 and 3.13 `sys.stdlib_module_names`, NOT the host's live set):
- `_resolve_python_module_path`: an absolute import whose top-level name is
a stdlib module no longer resolves to a loose same-named `.py` file or
namespace dir. A deliberate package shadow (`logging/__init__.py`) still
resolves — only loose modules/namespaces are refused.
- `_repoint_python_sibling_imports`: stdlib-named loose siblings are no
longer indexed as repoint targets.
- build.py #1504 pre-migration stem-alias index: a stdlib-named Python file
no longer claims the collapsed bare-stem alias ("logging"), which would
otherwise re-capture the external edge. The directory-scoped alias
("scripts_logging") is still registered, so genuine stale-id references to
the file keep healing.
Adds a stdlib-shadow regression plus a non-stdlib positive control.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
`graphify opencode install` wrote a plugin with only a named export, the
OpenCode 1 shape. OpenCode 2 rejects it ("Plugin must export a default
definition with an id and an effect or setup function") and reads the bare
path registered in opencode.json as an npm package name, so it fails twice
on every start.
The plugin is now one default export with `id`, `setup()` for OpenCode 2
and `server()` for OpenCode 1 (1.3.4 and later). The installer no longer
writes the opencode.json entry, because both runtimes discover
.opencode/plugins/ on their own, and it removes the entry older releases
wrote.
OpenCode loads .opencode/plugins/ from the project and from every directory
above it, and OpenCode 2 fails a second plugin with the same id ("Duplicate
plugin ID"). An install in ~ and one in a project are both normal, so the id
carries the file's location and the copies share one record of who was
reminded: both load, and a command still gets a single reminder.
Upgrading graphify does not touch a plugin already on disk. So the CLI now
rewrites an old-shape plugin in the working directory or in the home
directory on its next run, and drops the stale entry there, next to the
stale-skill refresh from #1805. It does so only when the `opencode` on PATH
reports 1.3.4 or later, because older OpenCode 1 releases cannot load the
new plugin and the one they have still works. The rewrite takes the same
per-directory lock as the skill refresh and replaces both files atomically,
because OpenCode watches them. GRAPHIFY_NO_AUTO_REFRESH=1 turns it off.
(cherry picked from commit 0e539e1f7a)
A file-scoped path endpoint that matched no symbol in that file fell through to
bag-of-words scoring, which then answered from a same-named symbol in another
file. Refuse like explain does (listing the files that do define the symbol)
instead of guessing across files; fixes the CLI path and the MCP shortest_path
tool via the shared resolver. Native Rust/C++ :: labels resolve via the exact
tier and are unaffected.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
A partially-symbolic backtick name (`%foo%`, S3 `[.cls`, dotted `my.op`)
normalized to the same _make_id suffix as an ordinary identifier, so the second
node was silently dropped by the seen_ids merge. Route every backtick-quoted
name through the reserved __operator__<hex> namespace, not only empty-suffix
operators; #4171's existing ids are unchanged. (The #4253 hang itself was
already fixed by #4171's resolve_local visited guard.)
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
File-scope begin/process/end/dynamicparam/clean parse as pseudo-commands
(command_name 'begin'), not named_block, so the command branch returned before
the child walk and any function/class defined inside was dropped. Descend into
those five block keywords' children; ids stay name-based and stable.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
An unqualified call fell through to a file-wide last-writer-wins label map, so
it bound to the last-declared same-named method. Walk the caller's class chain
(own class + in-file bases nearest-first, then enclosing classes) and bind a
single hit; fail-closed on a tie; free functions/constructors still resolve.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Resolve a.b.C(), xs[i].M(), and calls inside constructor bodies: a fail-closed
chain/element-type resolver (one declared type per hop, collection element only
for T[]/List<T>, bail on ambiguity) plus a ctor-body walk attributing calls to
the type node. Reuses the existing C# scoping; avoids same-pair method/calls
duplicates.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
