feat(editor): complete database, schema, and table segments in qualified names#1581
Conversation
…ied names Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
|
All contributors have signed the CLA ✍️ ✅ |
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: f6feb14f0e
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
| if let tableName = await schemaProvider.resolveAlias(dotPrefix, in: context.tableReferences) { | ||
| items = await schemaProvider.columnCompletionItems(for: tableName) | ||
| let schema = context.tableReferences.first { | ||
| $0.tableName.caseInsensitiveCompare(tableName) == .orderedSame | ||
| }?.schema |
There was a problem hiding this comment.
When two schemas contain the same table name and both are referenced with aliases (for example FROM s1.users u1 JOIN s2.users u2 ... WHERE u2.|), resolveAlias("u2") returns only "users", and this lookup then takes the schema from the first users reference. That makes completions for u2. fetch s1.users columns instead of s2.users, so schema-qualified alias columns are wrong in this common duplicate-table-name case; the resolution needs to keep the matched TableReference (or at least its schema) rather than re-finding by table name.
Useful? React with 👍 / 👎.
| await provider.setNamespaces( | ||
| schemas: services.schemaService.schemas(for: connectionId), | ||
| databases: currentDb.map { [$0] } ?? [] | ||
| ) |
There was a problem hiding this comment.
This only registers the current active database as a known database, even though loadSchema() has just loaded the full database list via DatabaseTreeMetadataService. In connections with multiple databases, SELECT * FROM OTHER_| cannot suggest any non-active database names, so the new database-segment completion advertised for Snowflake/BigQuery is limited to one database; use the loaded database metadata here instead of currentDb.map { [$0] }.
Useful? React with 👍 / 👎.
|
I have read the CLA Document and I hereby sign the CLA. |
…ame autocomplete with tests
datlechin
left a comment
datlechin
left a comment
There was a problem hiding this comment.
This look so good. Thank you
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: f694518a7c
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
| if let tableName = await schemaProvider.resolveAlias(dotPrefix, in: context.tableReferences) { | ||
| items = await schemaProvider.columnCompletionItems(for: tableName) | ||
| let schema = context.tableReferences.first { | ||
| $0.tableName.caseInsensitiveCompare(tableName) == .orderedSame | ||
| }?.schema | ||
| items = await schemaProvider.columnCompletionItems(for: tableName, schema: schema) |
There was a problem hiding this comment.
When the schema segment also matches a loaded table name, typing a qualified table name such as FROM sales.| resolves sales through resolveAlias first and returns columns for the sales table instead of tables in the sales schema. This makes the new schema-segment completion fail for common names like sales/analytics whenever a table with the same name has cached/fetchable columns; in namespace contexts the known schema/database branch needs to win before falling back to table/alias column completion.
Useful? React with 👍 / 👎.
2 participants
Summary
Improves SQL autocomplete for connections that organize objects as database → schema → table (BigQuery today, Snowflake with #1580):
SchemaService.runHierarchicalLoadpublishes.loaded([])and per-schema tables never reachedSQLSchemaProvider); it now receives the union of all loaded per-schema tables plus schema and database namesANALYTICS_PR→ database,ANALYTICS_PROD.→ schemas,ANALYTICS_PROD.DBT_MARTS.→ tables — including schemas never expanded in the sidebar, fetched on demand through the metadata poolFROM db.schema.orders o WHERE o.previously fetched columns from the active schema and came back empty;TableReferencenow carries the schema segment end to endSAFE_CASTon BigQuery,JSONB_AGGon PostgreSQL)No behavior change for flat-schema connections beyond the dialect functions appearing in completions.
Testing
testProviderAcceptsDatabaseType/testMySQLProviderTypeswhen suites run together reproduce identically onmainand predate this PRswiftlint lint --strictclean on all changed filesChecklist
CHANGELOG.mdupdated under[Unreleased]docs/features/autocomplete.mdx)Related: #1420, #1580
🤖 Generated with Claude Code