Add sqllogictests for scans after a schema change with no write since

Open
  • Is something wrong?
#75 0 comments 0 reactions 0 assignees View on GitHub

Maintainers usually reply within 1 day

Nobody has claimed this yet.

Assessment

Difficulty
4/5
Estimated time
3-5 days
Newbie friendliness
35/100
Issue type
Feature
Clarity
Mostly clear
Activity status
Active
Tech stack
rust, sql
Domain
databases

Research direction

Start with the sqllogictest files in the datafusion-iceberg repo and the engine setup that runs DataFusion only, then find where a test could call Transaction::update_schema() in iceberg-rust. Confirm the dependency includes apache/iceberg-rust#2904 before writing tests. Done means sqllogictests cover column add, drop, drop-and-re-add with the same name, and rename after a schema change with no write, and they pass.

Written by the indexing model from the issue text.

Description

Is your feature request related to a problem or challenge?

apache/iceberg-rust#2904 changes TableScanBuilder::build so that a scan that doesn't pin a snapshot projects the table's current schema, not the schema of the snapshot it reads. Before it, a scan of a table whose schema changed after its last write resolved columns against the older schema. That is the root cause of #42.

That PR checks the snapshot and schema a scan resolves with unit tests. End-to-end checks of the rows read back were asked to live here as sqllogictests rather than as integration tests in iceberg-rust (review comment).

Describe the solution you'd like

Once this crate depends on an iceberg-rust release that includes apache/iceberg-rust#2904, add sqllogictests that change a table's schema after its last write, with no write since, and check what queries return:

  • Column added: SELECT * returns the new column as NULL for every existing row. WHERE c IS NULL matches every row, and WHERE c = <value> matches none. This covers #42.
  • Column dropped: selecting the dropped column fails with a column-not-found error, and SELECT * no longer returns it.
  • Column dropped and added back under the same name: the new column reads as NULL, not the old column's values. Data files are matched by field ID, and the new field ID isn't in any file.
  • Column renamed: the values read back under the new name.

The sqllogictest engine here runs only DataFusion, and DataFusion SQL can't change an Iceberg table's schema. These tests need a way to do it, for example a setup step that calls Transaction::update_schema() or a Spark engine, as the README lists.

Additional context
  • #42: SELECT * fails after a column is added until the next write to the table.
  • #57 is related but not covered here: a provider built before the schema change keeps its cached schema.
Dominant language
Rust
Stars
30
Forks
30
Avg merge
3d 19h
Merged PRs (30d)
15

Getting set up

This project ships no dev container, Dockerfile or contributing guide, so setting up is up to you: start from its README, and see our first-contribution guide for the general steps.

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

More from apache/datafusion-iceberg

All issues in apache/datafusion-iceberg

Similar issues

More Rust issues

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.