apache / apache/datafusion

Fuzz underlying physical types for logical types in sqllogictests

Open
#17,947 2 comments 2 reactions 0 assignees View on GitHub
enhancement
Dominant language
Rust
Stars
9.3k
Forks
2.4k
Avg merge
3d 7h
Merged PRs (30d)
344

Description

### Is your feature request related to a problem or challenge?

If we have a function that accepts a string argument, this could be provided as `Utf8`, `LargeUtf8`, `Utf8View`, or via a `Dictionary` type; In SLT tests we would have to ensure we cover all these input types to ensure we don't accidentally miss an implementation.

### Describe the solution you'd like

Have a way in our SLT tests to fuzz the physical types of logical types; this extends to lists (`List`, `LargeList`, `FixedSizeList`) and other types too. Would need to consider tests which check the output type as well since those could be dependent on the input type.

### Describe alternatives you've considered

_No response_

### Additional context

Spawned from comment here: https://github.com/apache/datafusion/issues/13408#issuecomment-3368478752

Contributor guide

Open the contributing guide

Research direction

Start by reviewing the existing sqllogictests (SLT) and the discussion linked from issue #13408. Define how logical types such as strings and lists can exercise their physical variants, including Dictionary types, and identify how output-type assertions should behave when inputs vary. Done means SLT coverage can fuzz these physical representations without invalidating intentional output checks.

Written by the indexing model from the issue text.

Assessment

Tech stack
rust, sql
Domain
testing-qa
Issue type
Feature
Difficulty
5/5
Estimated time
Over a week
Activity status
Stale
Clarity
Mostly clear
Newbie friendliness
35/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.