Uh oh!
There was an error while loading. Please reload this page.
docs: sqlanvil docs site sourcing & build plan - #8
Closed
ihistand wants to merge 20 commits into
Closed
Conversation
Add three planning docs for the dataform → sqlanvil work: - CLAUDE.md: project guidance for Claude Code sessions - docs/postgres_first_class_design.md: spec for Postgres + Supabase adapters (replaces antigravity assessment's Phase 3+ framing) - docs/rename_checklist.md: exhaustive rename surface map Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
- All 8 .proto files: package dataform → package sqlanvil - Java options: com.dataform.protos → com.sqlanvil.protos - Go options: github.com/dataform-co/dataform → github.com/ihistand/sqlanvil - Field rename (wire-breaking): dataform_core_version → sqlanvil_core_version (in configs.proto + core.proto) - WORKSPACE name: df → sa - Bazel target //protos:dataform_proto → //protos:sqlanvil_proto cloud.google.com/dataform/docs/... URL comments retained — factually correct references to BigQuery partitioning docs. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
…te dead infra - git mv packages/@dataform → packages/@SQLAnvil (3 subdirs: cli, core, package.layer.json; all symbol-pinned package_name fields updated) - Sed @dataform/ → @sqlanvil/ across all BUILD files, .bzl, scripts, .ts/.js/.sqlx test fixtures (40+ files) - Root BUILD: gazelle:prefix → github.com/ihistand/sqlanvil - .github/CODEOWNERS: dataform-co/dataform-reviewers → ihistand - scripts/run: build target → //packages/@sqlanvil/cli:bin - Test fixture emails: allusers@dataform.co → allusers@example.com (consistent across constants.js, .sqlx fixtures, and matching assertions in tests/api/projects.spec.ts) Deletions (dead infra targeting upstream's GCP project): - cloudbuild-publish.yaml, cloudbuild-test.yaml, cloudbuild-version.yaml (Google Cloud Build configs for dataform-co's project) - scripts/publish, scripts/update_test_credentials (publishing to upstream npm + KMS key access against dataform-co's project) - test_credentials/{BUILD,bigquery.json.enc} + all //test_credentials:bigquery.json refs from 4 BUILD files (integration tests against BQ will need fresh credential wiring) NOT renamed (preserved historical/upstream references): - contributing.md upstream issue tracker links - readme.md links to Google Cloud Dataform docs - .bazelrc commented-out remote cache (already disabled) - cli/credentials.ts docs link (TODO: update once sqlanvil docs exist) - All `dataform.*` proto namespace usages in TS — deferred to phase C Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
…efaults
Path mapping:
- tsconfig.json paths: df/* → sa/*
- All `from "df/..."` / require("df/...") / import("df/...") → sa/* (450+)
- Runtime path strings in jit_loader.js + webpack configs + cli/index_test_base.ts
Proto namespace cascade (proto package dataform → sqlanvil triggers TS
generation under `sqlanvil` namespace):
- import { dataform, google } → import { sqlanvil, google } (and reverse)
- dataform.IFoo / dataform.JFoo / dataform.DbAdapter / etc. → sqlanvil.*
- camelCase field name dataformCoreVersion → sqlanvilCoreVersion
- Internal function name dataformCompile → sqlanvilCompile
Framework runtime defaults:
- globalAny.dataform → globalAny.sqlanvil (the user-facing sqlx namespace)
- __dataform_current_file → __sqlanvil_current_file (private global)
- "_dataform" table folder root default → "_sqlanvil"
- "dataform_assertions" assertion schema default → "sqlanvil_assertions"
- defaultSchema/defaultDatabase/defaultProject "dataform" → "sqlanvil"
- BigQuery job ID prefix "dataform-" → "sqlanvil-" (for BQ telemetry attribution)
- Test fixture labels team:"dataform" → team:"sqlanvil"
Doc comment anchors: configs#dataform-ActionConfig-* → configs#sqlanvil-*
(reference the docs site anchor IDs which derive from proto package name)
Test fixture strings: dataform-iceberg-test/-partial, dataform-database,
dataform-conn_id, dataform-integration-tests, dataform-test-schema-suffix
all renamed for consistency.
Preserved (will address in later phases or kept as upstream attribution):
- dataform.json strings (Phase D deletes the code path)
- dataform-co/dataform URLs (legitimate upstream attribution)
- dataform-open-source GCP project (test_credentials already deleted;
these integration tests are broken regardless)
- CLI help text "dataform [command]" (Phase E renames yargs scriptName)
- "invalid dataform json throws error" test name (Phase D deletes test)
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>Upstream Dataform marked dataform.json deprecated in favor of
workflow_settings.yaml. sqlanvil completes the cleanup — no
sqlanvil.json parallel, no back-compat fallback.
Removed code:
- core/workflow_settings.ts: readWorkflowSettings() dataform.json fallback
and "deprecated alongside" error path (workflow_settings.yaml is the
only source of truth now)
- cli/index.ts: dataformJsonPath presence check in projectDirMustExistOption
("does not appear to be a dataform directory" → "sqlanvil directory")
- cli/vm/compile.ts: global.dataformJson VM-time require
- cli/api/commands/init.ts: dataform.json existence check + init error
message ("Cannot init dataform project" → "Cannot init sqlanvil project")
- testing/run_core.ts: VM dataformJson global injection +
VALID_DATAFORM_JSON constant
- core/main_test.ts: 5 test blocks exercising the legacy path
("a valid dataform.json", "fails when both...", "invalid json file",
"contains unknown fields", "variables must be strings")
- tests/api/projects.spec.ts: "invalid dataform json throws error" test
- cli/index_compile_test.ts: replaced dataform.json fixture write with
equivalent workflow_settings.yaml
Removed fixtures:
- tests/api/projects/invalid_dataform_json/ (entire directory)
- tests/api/BUILD: invalid_dataform_json deps
Migrated fixtures dataform.json → workflow_settings.yaml:
- tests/integration/postgres_project/
- tests/api/projects/never_finishes_compiling/
- tests/api/projects/common_v2/
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>… sqlanvil
CLI (cli/index.ts):
- yargs scriptName: dataform → sqlanvil (sets program name in --help)
- All command descriptions: "Compile the dataform project" → "sqlanvil"
- Error messages: "dataform encountered an error" → "sqlanvil"
- "for Dataform to use" → "for sqlanvil to use"
- projectDirOption describe: "Dataform project directory" → "sqlanvil"
- jobPrefix template: dataform-${...}- → sqlanvil-${...}-
Help test (cli/index_help_test.ts): assertion strings match new help output.
Init test (cli/index_init_test.ts): defaultAssertionDataset literal + error
messages.
cli/credentials.ts: replaced broken docs link
(docs.sqlanvil.co/dataform-cli — which didn't exist on either domain) with
Google Cloud's official iam.docs.google.com link to service-account-key
creation.
VSCode extension (vscode/):
- extension.ts: language server ID + display name, command IDs
(dataform.compile → sqlanvil.compile), getConfiguration("sqlanvil")
- server.ts: spawn("sqlanvil"), all error message text
- package.json: rewrote manifest — name/publisher/displayName/repository/
icon/configuration namespace/command IDs all switch to sqlanvil. Version
reset 0.0.16 → 0.0.1 (fresh extension).
- Renamed vscode/dataform_logo.png → vscode/sqlanvil_logo.png (placeholder;
actual icon needs design before publishing)
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>- IDataformExtension → ISqlanvilExtension (and all imports/implementations: core/extension.ts, core/index.ts, core/main.ts, packages/sample-extension, packages/@sqlanvil/core) Docs: - readme.md: complete rewrite for sqlanvil positioning (was upstream Dataform copy referencing GCP-managed Dataform) - contributing.md: rewrite for sqlanvil — removed dataform-co/dataform upstream links, dropped GCP-credentials-team section (no longer accessible), added Postgres integration test prerequisite (Docker) - docs/configs-reference.md: placeholder until new proto schemas merge - docs/packages.md: light update (cloned base repo + framework name) - docs/postgres_reintegration_assessment.md: marked SUPERSEDED with pointer to postgres_first_class_design.md NOTICE: Apache 2.0 attribution file. Required by §4 of the license when distributing derivative works. Credits Dataform Co + Google LLC as upstream authors. Preserved upstream references (legitimate attribution / external packages): - readme/docs links to cloud.google.com/dataform (Google's product docs) - docs/packages.md sample package URLs (external community packages authored against upstream Dataform; not maintained by sqlanvil) - LICENSE (Apache 2.0 text, unchanged) Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
… dead infra
Comprehensive `Dataform` (capital D) → `sqlanvil` across all remaining
source code via brute-force sed on TS/proto files (BSD sed's word-boundary
regex didn't fire on first pass; explicit pattern caught the rest).
Auto-generated docs (docs/reference/*.md):
- Anchor IDs: <a name="dataform-..."> → <a name="sqlanvil-...">
- Cross-ref links: configs#dataform-ActionConfig-* → configs#sqlanvil-*
- Body text describing framework concepts: "Dataform" → "sqlanvil"
CLI test fixtures (cli/util_test.ts, cli/index_run_e2e_test.ts):
- jobId prefix "dataform-915a03fe1" → "sqlanvil-..."
- schema "dataform_test_schema_suffix" → "sqlanvil_test_schema_suffix"
- --job-labels team=dataform → team=sqlanvil
- uniqueDataset prefix dataform_e2e_osc_ → sqlanvil_e2e_osc_
cli/api/dbadapters/execution_sql.ts + matching goldens (*.sql):
- Generated DECLARE/SET variable name dataform_columns → sqlanvil_columns
(BQ on-schema-change synchronize/extend/fail paths)
cli/api/commands/init.ts: workflow_settings.yaml init writes
"defaultDataset: sqlanvil" (was "dataform")
cli/index_test_base.ts:
- DEFAULT_DATABASE: dataform-open-source → your-bigquery-project
- DEFAULT_RESERVATION reservation name dataform-test → sqlanvil-test
cli/vm/compile.ts + cli/vm/jit_worker.ts: node_modules/@dataform/core
bundle.js lookup → @sqlanvil/core (CLI tells users to run
`sqlanvil install`, not `dataform install`)
common/protos/index.ts: error messages now point at
github.com/ihistand/sqlanvil docs + issues, not dataform-co.
protos/configs.proto: "dataformCoreVersion" doc string + comments
updated.
tests/integration/bigquery_project + bigquery.spec.ts + examples/:
- defaultProject: dataform-open-source → your-bigquery-project
- 3-part BQ names rewritten with the placeholder
- Integration tests will need fresh credentials configured before they
can run (already noted in contributing.md)
- examples/.../snowy_repository_creation.ipynb: dataform_examples dataset
reference → your_examples_dataset
VSCode (vscode/{README.md,workflow_settings_yaml.schema.json,package-lock.json}):
- Deprecation notice referencing upstream extension removed
- JSON schema field dataformCoreVersion → sqlanvilCoreVersion
- package-lock.json name field
Deletions (dead infra):
- scripts/decrypt_secret (KMS keyring access to dataform-co's GCP project)
- scripts/create_gh_pr (sets origin URL to dataform-co/dataform.git)
Preserved legitimate references:
- protos/configs.proto: cloud.google.com/dataform/docs/partitions-clusters
URL comments (factual BigQuery partition/cluster reference docs)
- core/main.ts, core/main_test.ts, api/utils/postgres.ts: TODO comments
and upstream issue/PR links for historical context
- .bazelrc: commented-out remote cache configuration (already disabled)
- NOTICE: Apache 2.0 attribution to Dataform Co + Google LLC
- readme.md, CLAUDE.md, docs/*: attribution phrases like "upstream Dataform",
"Dataform OSS by Dataform Co"
- docs/reference/configs.md table rows referencing partition-clusters docs
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>The pinned Bazel 5.4 + 2022-era protobuf chain inherited from upstream does not build on macOS Tahoe + Apple Silicon: - Bazel 5.4 ships wrapped_clang without LC_UUID load commands; rejected by current dyld. - Bumping to Bazel 6.x triggers @bazel_tools//platforms removal failures in the WORKSPACE. - Bypassing wrapped_clang via system clang gets further but hits Apple SDK <-> old protobuf preprocessor conflicts (_stdio.h sprintf redefinition). Linux is unaffected. Until a future Bazel-7 + Bzlmod modernization PR lands, macOS contributors should build inside the dev container: ./scripts/docker-bazel build //protos:sqlanvil_proto ./scripts/docker-bazel test //core/... ./scripts/docker-bazel # interactive shell Dockerfile.dev: - Base: node:20-bookworm (gives node + npm) - Adds JDK 17, python 3, build-essential, zlib1g-dev, git, zip/unzip - Installs Bazelisk globally via npm - Mounts /root/.cache/bazel + /root/.cache/bazel-disk as named volumes for persistent Bazel cache across runs scripts/docker-bazel: - Lazy-builds the sqlanvil-dev image on first run - Mounts the repo at /workspace - Mounts named volumes sqlanvil-bazel-cache and sqlanvil-bazel-disk - Auto-attaches a TTY when stdin is a terminal (CI-safe otherwise) - No args → drops into bash inside the container contributing.md: documents the macOS compatibility caveat and the Docker workflow. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
When running bazel build //protos:sqlanvil_proto directly on the proto_library target (rather than going through ts_proto_library which wraps protoc with different flags), two pre-existing upstream bugs surface: 1. protos/configs.proto and core.proto import "extension.proto" without the protos/ prefix. Without strip_import_prefix on the proto_library, protoc passes -Iprotos/extension.proto=protos/extension.proto and the bare-name import fails. Adding strip_import_prefix = "/protos" removes the protos/ prefix so file references match. 2. jit.proto imports google/protobuf/timestamp.proto but the proto_library only declared deps on empty_proto and struct_proto. Added timestamp_proto. Both are upstream Dataform bugs (same content in upstream/main); they weren't caught because the upstream CI path went through ts_proto_library which has its own protoc invocation. Direct proto_library builds always would have failed. Verified working via: ./scripts/docker-bazel build //protos:sqlanvil_proto → Build completed successfully, 4 total actions This proves the dataform → sqlanvil rename is structurally sound at the proto layer (workspace name `sa`, package `sqlanvil`, target `sqlanvil_proto`, java/go options all resolve and protoc accepts them). Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Captures branch state, verification results, known pre-existing issues blocking fuller verification, and the recommended sequence of follow-up work (refresh node pin, then Postgres adapter PR, then Bazel modernization later). Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Rename sweep missed these 4 BUILD files that referenced the old `@df` workspace name in load() statements and dep labels. Without this fix, //core/... fails to analyze with: error loading package 'common/errors': Unable to find package for @df//testing:index.bzl: The repository '@df' could not be resolved Found while verifying //core/... in Docker.
Rename sweep missed: - packages/@sqlanvil/core/webpack.config.js — webpack resolve.alias - packages/sample-extension/webpack.config.js — webpack resolve.alias - packages/rollup.config.js — importsToBundle pattern - BUILD — ts_library `modules-fix` module_name Without these, //core/... fails to bundle: Module not found: Error: Can't resolve 'sa/core' in '...' Webpack alias still pointed `df` → RUNFILES/df, so any import of `sa/core` (post-rename TS path mapping) couldn't be resolved. Found while verifying //core/... in Docker.
The pinned Node v24.13.0 in WORKSPACE had stale SHA256 hashes that no longer matched nodejs.org's served binaries: Expected: e798599612f4bb71333a3397ab0d095fd62214e115aea45aa858a145fc72d67e Actual: aa881151bd0f9f154a0424dd60a72e9ce10672619121658c278a24327ef46831 Bumped to Node v20.20.2 LTS (matches the node:20-bookworm base in Dockerfile.dev). Fresh SHA256 hashes pulled from nodejs.org/dist/v20.20.2/SHASUMS256.txt for all 5 platforms. This was a stale-pin issue (not a Node 24 problem) — nodejs.org appears to have re-published the v24.13.0 archives at some point. Pinning to an LTS line is more stable. Verified in Docker: ./scripts/docker-bazel build //protos:ts → success ./scripts/docker-bazel build //core/... → success ./scripts/docker-bazel test //core/... → 16/16 pass Full tree (//...) still blocked by a separate pre-existing upstream issue: tools/postgres/BUILD references @io_bazel_rules_docker which is not declared in WORKSPACE.
…age directly Upstream Dataform removed @io_bazel_rules_docker from WORKSPACE in commit 8be595a (Mar 2024), but the restored Postgres adapter files in tools/postgres/BUILD still loaded `container_image` from it, breaking `bazel build //...`: error loading package 'tools/postgres': Unable to find package for @io_bazel_rules_docker//container:image.bzl: The repository '@io_bazel_rules_docker' could not be resolved Two paths considered: 1. Restore @io_bazel_rules_docker decl in WORKSPACE (~67 lines of deprecated bazel infra to maintain). 2. Refactor the fixture to use docker run directly. Chose (2). Modern Postgres LTS, no deprecated bazel rules, simpler dev flow. PostgresFixture now starts `postgres:15-alpine` via `docker run` on the local daemon — no Bazel image staging needed. Also rename DOCKER_CONTAINER_NAME from "postgres-df-integration-testing" to "postgres-sa-integration-testing" (rename sweep miss). Verified in Docker: ./scripts/docker-bazel build -- //... -//tools/postgres/... → 332 actions, Build completed successfully tools/postgres/... itself still needs `pg` + `pg-query-stream` added to package.json — that's Phase 1 of docs/postgres_reintegration_assessment.md (separate PR).
Phase 1 of `adapter/postgres-first-class` per docs/postgres_first_class_design.md §6. Adds the npm packages the restored Postgres adapter needs: - pg ^8.11.3 — node-postgres client - pg-query-stream ^4.5.3 — streaming query results - uuid ^9.0.0 — used by tools/postgres fixture - @types/pg ^8.11.0 — types for pg - @types/uuid ^9.0.0 — types for uuid Also removed @types/pg-query-stream from tools/postgres/BUILD; that package is a stub (pg-query-stream 4.x bundles its own types) and yarn warns when it's installed. Dropped from package.json too. yarn.lock regenerated against the new Node 20.20.2 LTS pin (PR #2). Verified in Docker: ./scripts/docker-bazel build //tools/postgres/... → success ./scripts/docker-bazel build //... → success (139 targets) ./scripts/docker-bazel test //... → 38/41 pass Remaining failures (not caused by this PR): - //cli:index_run_e2e_test — needs BigQuery test creds - //tests/integration:bigquery.spec — needs BigQuery test creds - //tests/api:projects.spec — pre-existing schema mismatch on `warehouse` property in workflow_settings validation Next phase (separate PR): relocate `api/dbadapters/postgres.ts` → `cli/api/dbadapters/postgres.ts` and fix imports.
… (Phase 2) Phase 2 of `adapter/postgres-first-class` per docs/postgres_first_class_design.md (Phases 1-2 inherited verbatim from docs/postgres_reintegration_assessment.md). Moves the restored Postgres adapter files into the modern Dataform/sqlanvil CLI layout: api/dbadapters/postgres.ts → cli/api/dbadapters/postgres.ts api/utils/postgres.ts → cli/api/utils/postgres.ts Updated imports (mechanical) to point at the current `cli/api/...` modules: sa/api/commands/credentials → sa/cli/api/commands/credentials sa/api/dbadapters/index → sa/cli/api/dbadapters/index sa/api/utils/postgres → sa/cli/api/utils/postgres sa/api/utils/results → sa/cli/api/utils/results sa/core/adapters → sa/cli/api/dbadapters/execution_sql Also dropped the (now-broken) imports for symbols that were deleted upstream when non-BQ warehouses were removed: sa/api/ssh_tunnel_proxy — gone, deleted via commit b4d2b2f sa/api/utils/error_parsing — gone (parseRedshiftEvalError) The file still references `SSHTunnelProxy` and `parseRedshiftEvalError` in the body; that's fine — file is **excluded from the BUILD glob** in both `cli/api/BUILD` and `cli/api/utils/BUILD` until Phase 3 does the proper interface alignment (executeRaw, deleteTable, schemas signature, sqlanvil.IJDBC → new PostgresConnection proto, SSH removal, error formatter swap). Verified: ./scripts/docker-bazel build //... → 139 targets, success Empty root-level api/ directory removed. Next: Phase 3a — implement IDbAdapter against the relocated file.
Adds the proto messages that the Postgres-first-class adapter needs, per docs/postgres_first_class_design.md §3 + §8.2: Action-level option blocks: - PostgresOptions — tablespace, fillfactor, unlogged, native declarative partitioning, btree/gin/gist/hash/ brin indexes, materialized view refresh policy - SupabaseOptions — Realtime publication, RLS enable, owner role, pgvector convenience config; nests PostgresOptions Top-level connection config: - BigQueryConnection — project, location, default_dataset - PostgresConnection — host, port, database, user, password, ssl_mode, default_schema (libpq-style) - SupabaseConnection — project_ref, service_role_key, default_schema, optional direct connection_string override - WarehouseConfig — discriminated union over the three connection variants via `oneof connection` This PR is **additive only**. Nothing references these messages yet — wiring into ActionConfig sub-messages (TableConfig, IncrementalTableConfig, etc.) and WorkflowSettings happens in subsequent phases: - Phase 3a PostgresDbAdapter consumes PostgresConnection - Phase 4 WorkflowSettings gets a `warehouse: WarehouseConfig` field; CLI dispatches on warehouse.kind - Phase 5 SupabaseDbAdapter + rlsPolicy/realtimePublication/wrapper/ vectorIndex action types consume SupabaseOptions Verified: ./scripts/docker-bazel build //protos:ts → success ./scripts/docker-bazel build //... → success (139 targets)
Rewrites cli/api/dbadapters/postgres.ts so it compiles against the
modern IDbAdapter interface, and wires the BUILD targets so the file
is part of the //cli/api library again (drops the Phase 2 excludes).
Interface alignment per docs/postgres_first_class_design.md §6 +
docs/postgres_reintegration_assessment.md §Phase 3:
- credentials sqlanvil.IJDBC → sqlanvil.IPostgresConnection
(new proto from Phase 3c)
- execute() returns IExecutionResult { rows, metadata }
— was raw any[]
- executeRaw() new, returns IExecutionResultRaw { rows,
schema, metadata } — schema kept empty (we
lean on the catalog query path for shape)
- withClientLock() callback receives an IDbClient with both
execute() and executeRaw() shimmed over the
underlying pg pool client
- schemas(database) now takes the database arg (ignored, since
Postgres is database-scoped), and filters
pg_catalog / pg_toast / pg_internal /
information_schema
- tables(database, schema) returns ITableMetadata[] (was ITarget[]) —
hydrates full metadata per row. Optional
schema filter is supported.
- deleteTable(target) new — transactional `drop table|view if exists
... cascade` keyed on the actual catalog type
- setMetadata() guards against null actualMetadata (table
doesn't exist yet) instead of crashing
Modernization (assessment doc §Phase 5):
- Drop SSHTunnelProxy block — Supabase + most managed Postgres serve
TLS on a public TCP port; SSH tunnel is no longer needed. The
SSHTunnelProxy class was deleted upstream in commit b4d2b2f.
- Drop parseRedshiftEvalError — replaced with parsePostgresEvalError in
cli/api/utils/error_parsing.ts. pg errors carry `.position` (byte
offset) but not line/column, so we preserve the message only.
- Drop preview() — wasn't in IDbAdapter.
SSL handling tightened:
ssl_mode = "disable" → no TLS
ssl_mode = "verify-ca"|"verify-full" → strict cert verification
ssl_mode = "" | "require" | "prefer" | "allow"
→ TLS with rejectUnauthorized=false
(matches default Supabase / RDS UX)
BUILD wiring:
- cli/api/BUILD: drop `dbadapters/postgres.ts` exclude; add `@npm//pg`
and `@npm//@types/pg` deps.
- cli/api/utils/BUILD: drop `postgres.ts` exclude; add `@npm//pg`,
`@npm//@types/pg`, `@npm//pg-query-stream` deps.
Verified:
./scripts/docker-bazel build //cli/api/... → 8 targets, success
./scripts/docker-bazel build //... → 139 targets, success
./scripts/docker-bazel test //... → 38/41 pass (same 3
pre-existing failures: 2× missing BQ creds, 1× projects.spec
`warehouse` validation issue — none regressed by this PR)
Next phase (Phase 3b): Postgres SQL generator under
core/compilation_sql/postgres/ — one generator per action type with
golden-output tests.Captures the legal + structural plan for sqlanvil's documentation site: - Tier 1 source: upstream/main:docs/ — Apache 2.0, already inherited - Tier 2 source: docs.cloud.google.com/dataform — CC BY 4.0 - Tier 3: do-not-pull list (cloud.google.com marketing, screenshots, etc.) License compliance worked out for both tiers — Apache 2.0 covered by extending the existing NOTICE; CC BY 4.0 needs per-page attribution + ATTRIBUTIONS.md. Phased doc roadmap aligned to code phases: - D1: pull Tier 1 source as-is, rebrand - D2: rewrite reference pages Postgres-first - D3: net-new Supabase action type pages - D4: quickstart + CLI ref adapted from Tier 2 - D5: architecture & concept pages (mostly original) Site stack recommendation: Astro Starlight on Vercel (docs.sqlanvil.com), sibling repo ../sqlanvil-docs/ so doc deploys don't trigger code CI. Trademark scrubbing (dataform → sqlanvil) treated as separate concern from copyright — same playbook as the code rename PR (#1).
ihistand
commented
Jun 7, 2026
CollaboratorAuthor
Superseded: the Postgres reintegration + Dataform→SQLAnvil rename landed on |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for freeto join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Captures the legal + structural plan for sqlanvil's documentation site at `docs/docs_site_plan.md`.
Source hierarchy
License footer on Tier 2 confirmed verbatim: "Except as otherwise noted, the content of this page is licensed under the Creative Commons Attribution 4.0 License, and code samples are licensed under the Apache 2.0 License."
Phased roadmap
Each phase aligned to a code phase from `postgres_first_class_design.md`.
Site stack recommendation
License compliance
Independent of code stack
This is a planning doc only. No code changes. Can merge anytime; doesn't block the postgres-first-class PR stack.
Test plan
🤖 Generated with Claude Code