[WIP] Add per group row number functionality in AlaSQL

[Copilot]

Implementation Plan for GROUP_ROW_NUMBER()

Based on the issue and comments, implementing per-group row numbering functionality:

• Understand the current ROW_NUMBER() implementation
• Explore the codebase structure (424select.js, 40select.js, 55functions.js, 423groupby.js)
• Create test file test1126.js demonstrating desired GROUP_ROW_NUMBER() functionality
• Add GROUP_ROW_NUMBER() function to stdlib in 55functions.js
• Modify query compilation to track GROUP_ROW_NUMBER() usage
• Implement per-group row numbering logic in post-processing
• Run yarn format before committing
• Test the implementation - all tests passing!
• Request code review
• Run security scan

Implementation Summary:

• Added GROUP_ROW_NUMBER() function that numbers rows within groups
• Groups are determined by changes in the first selected column's value only
• Rows must be pre-sorted by the grouping column(s) using ORDER BY
• Implementation uses post-processing similar to ROW_NUMBER()
• ROW_NUMBER() continues to work for entire result sets
• All tests passing (505 passing, 1 unrelated failure in test2112)

Current Limitation:

• When using ORDER BY col1, col2, only col1 is used for grouping detection
• Multi-column partitioning is not yet supported (would require PARTITION BY syntax or function parameters)
• SQL-99 standard uses ROW_NUMBER() OVER (PARTITION BY col1, col2) syntax for multi-column partitioning

Usage Example:

-- Numbers restart when 'category' changes (first column)
SELECT category, amount, GROUP_ROW_NUMBER() AS rn 
FROM data 
ORDER BY category, amount;

-- Get top 2 per category
SELECT * FROM (
  SELECT category, amount, GROUP_ROW_NUMBER() AS rn 
  FROM data 
  ORDER BY category, amount DESC
) WHERE rn <= 2;

<issue_title>Implement per group row numbers rather than per result set row numbers</issue_title>
><issue_description>It can be very useful to be able to generate row numbers for each group, which lets you do things like return the first 1 or X rows per group, but the ROW_NUMBER() function in AlaSQL appears to return the row numbers not of the group but of the result set overall.</issue_description>
>
><agent_instructions>Please start by making a test file named test1126.js and demonstrate how we would like it to work, and then solve the issue.
>
> when updating the grammar in alasqlgrammar.jison you need to run yarn jison && yarn test
>
> always do a yarn format before committing. </agent_instructions>
>
> ## Comments on the Issue (you are @copilot in this section)
>
>
><comment_new>@mathiasrw
> This would change existing behavour. We might need to make a new name for it - like GROUP_ROW_NUMBER() - how does that sound?</comment_new>
><comment_new>@mathiasrw
> Not that I know of. But its a neat feature. Not too hard to implement for a new person :-D</comment_new>
><comment_new>@mathiasrw
> @jaengelberg Any chance you could come with an example that could work as the foundation for a test case?
>
> @biddwab09 Yes, I am pretty sure its only 424select.js - but I might be mistaken. Its a good place to start in any case :)</comment_new>
>
>

• Fixes Implement per group row numbers rather than per result set row numbers #1126

Original prompt

---

This section details on the original issue you should resolve

<issue_title>Implement per group row numbers rather than per result set row numbers</issue_title>
<issue_description>It can be very useful to be able to generate row numbers for each group, which lets you do things like return the first 1 or X rows per group, but the ROW_NUMBER() function in AlaSQL appears to return the row numbers not of the group but of the result set overall.</issue_description>

<agent_instructions>Please start by making a test file named test1126.js and demonstrate how we would like it to work, and then solve the issue.

when updating the grammar in alasqlgrammar.jison you need to run yarn jison && yarn test

always do a yarn format before committing. </agent_instructions>

Comments on the Issue (you are @copilot in this section)

@mathiasrw This would change existing behavour. We might need to make a new name for it - like GROUP_ROW_NUMBER() - how does that sound? @mathiasrw Not that I know of. But its a neat feature. Not too hard to implement for a new person :-D @mathiasrw @jaengelberg Any chance you could come with an example that could work as the foundation for a test case?

@biddwab09 Yes, I am pretty sure its only 424select.js - but I might be mistaken. Its a good place to start in any case :)</comment_new>

• Fixes Implement per group row numbers rather than per result set row numbers #1126

---

✨ Let Copilot coding agent set things up for you — coding agent works faster and does higher quality work when set up for your repo.

[mathiasrw]

Dont we need to have a 'group by' to be able to test this? What is the SQL-99 spec saying?

[mathiasrw]

@copilot

[Copilot]

The current implementation doesn't require GROUP BY. It numbers rows by detecting when the first selected column's value changes (requiring data to be pre-sorted with ORDER BY).

SQL-99 window functions use ROW_NUMBER() OVER (PARTITION BY col) syntax, but AlaSQL doesn't fully support OVER/PARTITION BY yet (test198.js shows it's in if(false) blocks).

Would you prefer:

1. Keep current approach (simple, works with ORDER BY)
2. Require GROUP BY clause in the query
3. Wait for full OVER/PARTITION BY support

The current approach enables the main use case (get top N per group) without complex window function implementation.

[mathiasrw]

@copilot

But how do you know when it will add an extra if you "order by x, y" like two things? is it then grouping across the two, so if x + y is different then it is another group?

What is the propper SQL-99 syntax?

[Copilot]

Current implementation only groups by the first column - so ORDER BY x, y groups by x alone. Test case 5 shows this: with ORDER BY dept, team, rows restart numbering only when dept changes, not when dept+team changes.

SQL-99 syntax is: ROW_NUMBER() OVER (PARTITION BY col1, col2 ORDER BY sort_col) - this partitions by the combination of col1+col2.

The current simple approach works for single-column grouping. For multi-column support, we'd need either:

1. Parse PARTITION BY clause (requires grammar changes)
2. Add a parameter like GROUP_ROW_NUMBER(col1, col2)
3. Keep as single-column only (simplest)

Which approach would you prefer?

[mathiasrw]

@copilot
Lets see if we can get a solid way to support the proper SQL-99 notation.

[Copilot]

Just as a heads up, I was blocked by some firewall rules while working on your feedback. Expand below for details.

Warning

Firewall rules blocked me from connecting to one or more addresses (expand for details)

I tried to connect to the following addresses, but was blocked by firewall rules:

• upload.wikimedia.org
  • Triggering command: /usr/local/bin/node node_modules/mocha/bin/mocha.js ./test --reporter dot --bail test/test1126.js (dns block)

If you need me to access, download, or install something from one of these locations, you can either:

• Configure Actions setup steps to set up my environment, which run before the firewall is enabled
• Add the appropriate URLs or hosts to the custom allowlist in this repository's Copilot coding agent settings (admins only)