entgql: schema-qualify M2M join table in totalCount loader by michaelcaulley · Pull Request #636 · ent/contrib

michaelcaulley · 2026-06-16T20:22:37Z

The gql_collection.go totalCount loader for M2M edges builds the join table via sql.Table(...) but never calls joinT.Schema(...). With the sql/schemaconfig feature enabled, this emits an unqualified join table in the generated SQL while the rest of the query is correctly schema- qualified, producing relation does not exist errors at runtime.

Hook the same dialect/sql/query/eagerloading/join/* template extension point that ent core's query.tmpl uses for the loader path, so the join table picks up its schema from schemaConfig. No-op when the feature is disabled (verified: all 6 integration tests regenerate byte-identically).

a8m · 2026-06-17T13:00:08Z

cc @chinhnguyenh for review too.

Copilot

Pull request overview

This PR fixes schema qualification for M2M join tables in the totalCount loader generated by entgql, ensuring queries work correctly when Ent’s sql/schemaconfig feature is enabled (avoiding runtime relation does not exist errors due to unqualified join tables).

Changes:

Hooks the dialect/sql/query/eagerloading/join/* template extension point into the M2M totalCount loader path.
Ensures the join-table builder (joinT := sql.Table(...)) can be modified by the same join-schema logic used by Ent core eager-loading queries.
Keeps behavior a no-op when no matching join templates are present (e.g., feature disabled).

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

chinhnguyenh · 2026-06-17T14:05:22Z

Look good to me.
It would be nice if you can add a small example project for it.

michaelcaulley · 2026-06-17T15:09:26Z

Look good to me. It would be nice if you can add a small example project for it.

Hi, @chinhnguyenh, are you on Discord? If so, can I reach out there? I have a few questions about the best way to add a test. If we want to create a new multischema project, we'll need MySQL to test it. Otherwise, we can limit the test to reading the generated code.

Edit: I've added each option as a separate commit. If you'd like to discuss either option, please feel free to message me on Discord.

The gql_collection.go totalCount loader for M2M edges builds the join table via `sql.Table(...)` but never calls `joinT.Schema(...)`. With the `sql/schemaconfig` feature enabled, this emits an unqualified join table in the generated SQL while the rest of the query is correctly schema- qualified, producing `relation does not exist` errors at runtime. Hook the same `dialect/sql/query/eagerloading/join/*` template extension point that ent core's query.tmpl uses for the loader path, so the join table picks up its schema from `schemaConfig`. No-op when the feature is disabled (verified: all 6 integration tests regenerate byte-identically). Co-authored-by: Isaac

Add a test (entgql/template_test.go) that renders the collection template with and without the sql/schemaconfig feature and asserts the M2M totalCount join table is schema-qualified only when the feature is enabled. Covers the fix from the preceding commit at the codegen level and runs in CI against SQLite.

Add internal/multischema, a minimal Group<->User M2M project with the sql/schemaconfig feature, plus a MySQL-backed functional test that places the join table in a separate schema and drives totalCount through the gqlgen + ent runtime. ent omits schema qualifiers for SQLite (see (*sql.Builder).writeSchema), so the schema-qualified join is only observable on a schema-aware database; the test skips when MySQL is unreachable. Add a mysql8 service to the CI test job (mirroring ent's integration CI) so the test runs instead of skipping, and the go-sql-driver/mysql dependency it requires. Co-authored-by: Isaac

chinhnguyenh · 2026-06-17T20:35:03Z

Look good to me. It would be nice if you can add a small example project for it.

Hi, @chinhnguyenh, are you on Discord? If so, can I reach out there? I have a few questions about the best way to add a test. If we want to create a new multischema project, we'll need MySQL to test it. Otherwise, we can limit the test to reading the generated code.

Edit: I've added each option as a separate commit. If you'd like to discuss either option, please feel free to message me on Discord.

We don't need to test it with MySQL, verifing the generated code is enough.
Please rename the project to todoschema to keep consistency with others.

chinhnguyenh · 2026-06-17T20:39:34Z

+// TestCollection_M2MTotalCountSchemaConfig verifies that the M2M totalCount
+// loader schema-qualifies its join table when the sql/schemaconfig feature is
+// enabled, and leaves it unqualified otherwise.
+func TestCollection_M2MTotalCountSchemaConfig(t *testing.T) {


Create gen_test.go in the example project and create tests there instead.

If we're testing only the generated code, do we need the new project? Can I drop my last commit?

The example project is what CI regenerates and diffs. The test only reads the output, but the project is what keeps that output committed, compilable, and in sync with the templates.

Anyway, it's just a nice-to-have.

* entgql: fix field arguments collection (ent#495) * entgql: collect interafce/union selection fields of unique edges (ent#498) * entproto: support enum values with special characters (ent#487) * build(deps): bump golang.org/x/tools from 0.7.0 to 0.9.1 (ent#504) Bumps [golang.org/x/tools](https://github.com/golang/tools) from 0.7.0 to 0.9.1. - [Release notes](https://github.com/golang/tools/releases) - [Commits](golang/tools@v0.7.0...v0.9.1) --- updated-dependencies: - dependency-name: golang.org/x/tools dependency-type: direct:production update-type: version-update:semver-minor ... Signed-off-by: dependabot[bot] <support@github.com> Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com> * entgql: support collecting fields of renamed types (ent#505) * entgql: implement interfaces and collect all node implementors (ent#507) * entgql: support generating non-optional Map fields (ent#508) * feat: add SkipGenFile option (ent#499) * entgql: fixed Paginate query for count (ent#512) * entgql: fixed Paginate query for count * Update entgql/template/pagination.tmpl Co-authored-by: Ariel Mashraki <7413593+a8m@users.noreply.github.com> * entgql: codegen --------- Co-authored-by: Ariel Mashraki <7413593+a8m@users.noreply.github.com> * entgql/internal/todo: add example for json as mutation inputs (ent#513) * entproto/cmd: fix setting target path (ent#521) * entproto: add support for int slices (ent#522) * entproto: support json for integer type slices * adding int slices to grpc test suite --------- Co-authored-by: jamslinger <80337165+jamslinger@users.noreply.github.com> * entgql: privatize columns used in cursors predicates (ent#537) * privatize columns used in cursors predicates * add tests * entgql: avoid naming conflicts in toCursor template (ent#538) * fix: rename variable when iterating through p.order to prevent conflict with $r that starts with "O" character * fix: run go generate on entgql * entgql: upgrade gqlgen and disable getters (ent#544) * entgql: upgrade gqlgen and disable getters * entgql: codegen * chore: fixed go.mod * add AllowClientUUIDs for client UUIDs in POSTs (ent#541) * chore: upgrade ent and atlas (ent#554) * chore: upgrade ent and atlas * fix: bump go version to 1.20 * .ci: upgrade golangci * chore: codegen * entgql: fixed go-fmt (ent#555) * all: upgrade ent and gqlgen to latest (ent#564) * entgql: ensure order fields are prefixed with type name (ent#565) Also when order field is set on an edge * entgql: add missing directive's argument (ent#567) * entgql: avoid using cte queries for single node paths (ent#569) No need to limit per node, when a single node is returned in a path * entgql: support skipping tx-opening by operation or field name (ent#571) * entgql: improve var name collision avoidance in toCursor (ent#572) Improves variable name collision avoidance in the `toCursor` method of the pager types in the `gql_pagination` template by suffixing all hard-coded variables names with an underscore. * fix tests * fix * go: update gqlgen (ent#583) * update gqlgen pkg * run code gen * update proto * gen protobuf * fix tests * update protoc version in ci * update back protoc version * entgql/template: improve goimports time by adding missing imports (ent#584) * ci: use node20 actions because node16 actions are deprecated (ent#588) * all: bump ent and atlas to latest version (ent#596) * all: bump ent and atlas to latest version * chore: update test due to msgpack upgrade * entgql: add annotation Deprecated for deprecated field (ent#595) * entgql: support SkipMutationCreateInput (and update) on entity (ent#598) * entgql/internal/examples: fix template rendering for pulid example (ent#599) * entgql: fixed zero value on `first`/`last` when receive `null` input (ent#601) * entgql: use dynamic Value/GetValue method name in pagination with edge order fields (ent#603) Updates to use the `ValueName` method to get the name of the dynamic `Value`/`GetValue` method name for a type instead of using the hard-coded name `Value` in the `pagination.tmpl` file. * go: use go-viper/mapstructure/v2 over archived mitchellh/mapstructure (ent#606) * entgql/templates: ability to add extra fields and methods to mutation inputs (ent#608) * ability to add additonal fields to mutation input * Update mutation_input.tmpl removed the model keyword * chore: bump go to 1.23.4 and gqlgen to v0.17.51 (ent#607) * chore: bump go to 1.23.4 and gqlgen to v0.17.51 gqlgen v0.17.52 breaks field collection see 99designs/gqlgen@f81239e * chore: bump golangci and go version * entgql: upgrade to 99designs/gqlgen@v0.17.62 and vektah/gqlparser/v2@v2.5.21 (ent#609) * Update github.com/99designs/gqlgen@v0.17.61 and github.com/vektah/gqlparser/v2@v2.5.21 Signed-off-by: Steve Coffman <steve@khanacademy.org> * Re-generate Signed-off-by: Steve Coffman <steve@khanacademy.org> * Re-generate with correct protobuf versions Signed-off-by: Steve Coffman <steve@khanacademy.org> * Try again with protoc v3.19.4 Signed-off-by: Steve Coffman <steve@khanacademy.org> * minor tweak to go.mod and avoid pager getting stuck Signed-off-by: Steve Coffman <steve@khanacademy.org> * Update gqlgen to v0.17.62 Signed-off-by: Steve Coffman <steve@khanacademy.org> * regenerate? Signed-off-by: Steve Coffman <steve@khanacademy.org> * Minor tweak Signed-off-by: Steve Coffman <steve@khanacademy.org> * re-re-generate Signed-off-by: Steve Coffman <steve@khanacademy.org> * Add tools.go and run go mod tidy Signed-off-by: Steve Coffman <steve@khanacademy.org> * Generate using nektos/act -j generate Signed-off-by: Steve Coffman <steve@khanacademy.org> * Go generate again Signed-off-by: Steve Coffman <steve@khanacademy.org> * Generate using nektos/act -j generate Signed-off-by: Steve Coffman <steve@khanacademy.org> * go clean --modcache and then go mod tidy Signed-off-by: Steve Coffman <steve@khanacademy.org> --------- Signed-off-by: Steve Coffman <steve@khanacademy.org> * entgql: bump gqlgen version (ent#610) * fix: regenerate resolvers on codegen regeneration is required so that new ent fields and schemas are included * chore: bump gqlgen to v0.17.63 and regenerate code * entgql: support global id feature from ent (ent#611) * entgql: bump entgo and update template for receiver logic (ent#617) * entgql: bump entgo and update template for receiver logic * entgql: codegen * entoas: codegen * entproto: codegen * schemast: codegen * .github/workflows: update ci.yml workflow * entgql: bump gqlgen to latest version (ent#618) * entgql: add the CollectedFor annotation (ent#623) * fix: remove duplicate graphql import in generated files (ent#629) * entgql: prevent panic in case of a missing operation context (ent#630) * entgql: expose TableName for sql/global is feature (ent#632) * entgql: support collecting multiple ent fields for GraphQL field (ent#633) * entgql: replace golang.org/x/exp/slices import with standard slices package (ent#634) * entgql: fix relay connection eager-load on union/interface fragments (ent#635) Walk nested connection selections from the connection field with the current type condition instead of from the resolver field context with nil satisfies, which merged inline fragments and caused wrong eager loads. * entgql: schema-qualify M2M join table in totalCount loader (ent#636) * entgql: schema-qualify M2M join table in totalCount loader The gql_collection.go totalCount loader for M2M edges builds the join table via `sql.Table(...)` but never calls `joinT.Schema(...)`. With the `sql/schemaconfig` feature enabled, this emits an unqualified join table in the generated SQL while the rest of the query is correctly schema- qualified, producing `relation does not exist` errors at runtime. Hook the same `dialect/sql/query/eagerloading/join/*` template extension point that ent core's query.tmpl uses for the loader path, so the join table picks up its schema from `schemaConfig`. No-op when the feature is disabled (verified: all 6 integration tests regenerate byte-identically). Co-authored-by: Isaac * entgql: add codegen test for M2M totalCount schema qualification Add a test (entgql/template_test.go) that renders the collection template with and without the sql/schemaconfig feature and asserts the M2M totalCount join table is schema-qualified only when the feature is enabled. Covers the fix from the preceding commit at the codegen level and runs in CI against SQLite. * entgql: add multischema project with MySQL test for M2M totalCount Add internal/multischema, a minimal Group<->User M2M project with the sql/schemaconfig feature, plus a MySQL-backed functional test that places the join table in a separate schema and drives totalCount through the gqlgen + ent runtime. ent omits schema qualifiers for SQLite (see (*sql.Builder).writeSchema), so the schema-qualified join is only observable on a schema-aware database; the test skips when MySQL is unreachable. Add a mysql8 service to the CI test job (mirroring ent's integration CI) so the test runs instead of skipping, and the go-sql-driver/mysql dependency it requires. Co-authored-by: Isaac * entgql: rename multischema project to todoschema, drop MySQL test (ent#637) --------- Signed-off-by: dependabot[bot] <support@github.com> Signed-off-by: Steve Coffman <steve@khanacademy.org> Co-authored-by: Ariel Mashraki <7413593+a8m@users.noreply.github.com> Co-authored-by: jamslinger <80337165+jamslinger@users.noreply.github.com> Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com> Co-authored-by: Zxilly <zxilly@outlook.com> Co-authored-by: Giau. Tran Minh <giau.tmg@gmail.com> Co-authored-by: Rotem Tamir <rotemtamir@gmail.com> Co-authored-by: Amr Tamimi <amrnt@users.noreply.github.com> Co-authored-by: Tantatorn Suksangwarn <timefear007@gmail.com> Co-authored-by: Giau. Tran Minh <12751435+giautm@users.noreply.github.com> Co-authored-by: A B <bartke@users.noreply.github.com> Co-authored-by: Henry Woody <github@henrywoody.com> Co-authored-by: Ronen Lubin <63970571+ronenlu@users.noreply.github.com> Co-authored-by: ASAKURA Kazuki <32762324+Arthur1@users.noreply.github.com> Co-authored-by: Tran Minh Luan <tranminhluan1206@gmail.com> Co-authored-by: Jeffrey Stiles <jbstiles@gmail.com> Co-authored-by: Kai Dinghofer <10863868+kdevo@users.noreply.github.com> Co-authored-by: Henry Woody <git@henrywoody.com> Co-authored-by: Justin Cummins <sul3n3t@gmail.com> Co-authored-by: Aman Bansal <khanakia@gmail.com> Co-authored-by: Michael Caulley <me@mikecaulley.com> Co-authored-by: Steve Coffman <StevenACoffman@users.noreply.github.com> Co-authored-by: Michael Caulley <michael.caulley@databricks.com> Co-authored-by: Jannik Clausen <12862103+masseelch@users.noreply.github.com> Co-authored-by: Noam Cattan <62568565+noamcattan@users.noreply.github.com> Co-authored-by: Chinh Nguyen <monkeynhucci@gmail.com> Co-authored-by: Claude Fable 5 <noreply@anthropic.com>

michaelcaulley force-pushed the entgql-schemaconfig-m2m-totalcount branch 2 times, most recently from 17ccd50 to 31b1a58 Compare June 16, 2026 20:59

a8m requested a review from giautm June 17, 2026 13:00

a8m requested a review from Copilot June 17, 2026 13:00

Copilot started reviewing on behalf of a8m June 17, 2026 13:00 View session

Copilot AI reviewed Jun 17, 2026

View reviewed changes

michaelcaulley force-pushed the entgql-schemaconfig-m2m-totalcount branch from 31b1a58 to 8defc40 Compare June 17, 2026 15:39

michaelcaulley added 2 commits June 17, 2026 11:40

michaelcaulley force-pushed the entgql-schemaconfig-m2m-totalcount branch from 8defc40 to d3a3d97 Compare June 17, 2026 15:42

michaelcaulley force-pushed the entgql-schemaconfig-m2m-totalcount branch from d3a3d97 to 659006e Compare June 17, 2026 15:59

chinhnguyenh reviewed Jun 17, 2026

View reviewed changes

a8m approved these changes Jun 18, 2026

View reviewed changes

a8m merged commit 6a207cd into ent:master Jun 18, 2026
4 checks passed

thejensemann mentioned this pull request Jul 4, 2026

Merge upstream ent/contrib master (2026) + dependency maintenance spread-ai/ent-contrib#19

Merged

Navigation Menu

Uh oh!

entgql: schema-qualify M2M join table in totalCount loader - #636

entgql: schema-qualify M2M join table in totalCount loader#636
a8m merged 3 commits into
ent:masterfrom
michaelcaulley:entgql-schemaconfig-m2m-totalcount

michaelcaulley commented Jun 16, 2026

Uh oh!

a8m commented Jun 17, 2026

Uh oh!

Copilot AI left a comment

Uh oh!

chinhnguyenh commented Jun 17, 2026

Uh oh!

michaelcaulley commented Jun 17, 2026 •

edited

Loading

Uh oh!

chinhnguyenh commented Jun 17, 2026

Uh oh!

chinhnguyenh Jun 17, 2026

Uh oh!

michaelcaulley Jun 17, 2026

Uh oh!

chinhnguyenh Jun 17, 2026

Uh oh!

chinhnguyenh Jun 17, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

Uh oh!

Conversation

michaelcaulley commented Jun 16, 2026

Uh oh!

a8m commented Jun 17, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

chinhnguyenh commented Jun 17, 2026

Uh oh!

michaelcaulley commented Jun 17, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

chinhnguyenh commented Jun 17, 2026

Uh oh!

chinhnguyenh Jun 17, 2026

Choose a reason for hiding this comment

Uh oh!

michaelcaulley Jun 17, 2026

Choose a reason for hiding this comment

Uh oh!

chinhnguyenh Jun 17, 2026

Choose a reason for hiding this comment

Uh oh!

chinhnguyenh Jun 17, 2026

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

michaelcaulley commented Jun 17, 2026 •

edited

Loading