Skip to content

fuzz with SQLancer #239

@jayshrivastava

Description

@jayshrivastava

See #222

We can generate SQL queries using SQLancer https://github.com/datafusion-contrib/datafusion-sqlancer and verify the queries run correctly in distributed datafusion (see property-based validators added in #231)

There's some work to do

  • Add arrow flight server for SQLancer to run on
    • Intercept CREATE TABLE and INSERT statements and modify them so we write data to testdata/sqlancer/data
    • Intercept SELECT statements and write them to testdata/sqlancer/queries. Note that queries still need to be executed so SQLancer can use the data to generate more queries (ex. queries on the information scheam table for sure - unclear if it uses normal SELECT queries)
    • Data + queries = fuzz input. We run this input against distributed datafusion and validate using property based tests
  • The instructions for sqlancer say to run it with java. We probably need to compile and save a linux sqlancer binary in this repo or in git lfs.

Note that sqlancer does not support nested queries and window functions at the time of writing.

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions