test(dla): 125-table large Postgres fixture + e2e findings report - #16
Merged
Conversation
…configs - new fixture at apps/dla/tests/fixtures/postgres_large (port 55433, container dla_fixture_postgres_large) with 5 schemas: star/snowflake regions, conformed dims, 9 junction tables, a no-FK staging zone, 25 generic distractor tables, text-heavy tables, and structural edge cases (reserved-word/quoted/60-char identifiers, composite and multi-column FKs, self-referencing FKs, 110-column wide table, 100k-row tall table, jsonb/uuid/array/enum/range types) - seeded data-quality ground truth: empty tables, all-null and constant columns, 70% null rate, inferred and declared-NOT-VALID broken FKs, orphaned staging joins, mixed-case status values - example configs: postgres_large.yaml (all schemas) and postgres_large_staging_only.yaml (inference-only slice) - fixture README documents every region and expected detection
Full results of running the complete offline pipeline against the 125-table fixture: per-command results, performance numbers, idempotency verdict, and 18 ranked defects (manifest overcount on multi-schema sources, jsonb/array profiling failures with no readiness issue, SIGINT swallowed by dla run, recommender junction-vs-text scoring tie behavior, broken-fk type coercion, exit-code deviations, inference and pattern-detector false positives).
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 free
to 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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
What
A second, much larger test fixture at
apps/dla/tests/fixtures/postgres_large/(docker-compose on port 55433, container name distinct from the small fixture — the two coexist), plus ready-to-run configs and the findings report from a full end-to-end pipeline run against it.Fixture contents (125 tables / 673 columns / 5 schemas / ~130k rows):
stagingschema (simulates cloud-warehouse dumps; relationship inference only)NOT VALIDbroken FKREADME.mddocuments every region and seeded issue;FINDINGS.mdis the e2e reportWhy
The L2 gap analysis (G4) showed the 15-table fixture cannot prove or disprove any retrieval claim — full-schema prompt-stuffing is indistinguishable from retrieval at that scale. This fixture is the precondition for L2's schema-linking eval, and running L1 end-to-end against it was the first real-scale exercise of the shipped pipeline.
Headline e2e results (full detail in FINDINGS.md)
dla run17.4s; bundle 13MB / 3,350 filesdla runswallows SIGINT · recommender scores vector(6) over knowledge_graph(4) on a junction-rich 125-table schema (density diluted by table count) · broken-fk check lacks type coercion (350 false orphans) · two exit-code deviationsNo production code is changed in this PR — fixture, configs, and findings only. Defect fixes can be triaged from FINDINGS.md into the hardening backlog.