Handle Multiple ORDER BYs with DEFAULTs

[Lightning11wins]

This PR will add functionality for handling multiple ORDER BY statements. Later statements append to earlier ones, or overwrite them if they use the DEFAULT keyword, as suggested by Greg.

This PR includes both docs and testing for the new functionality and the DEFAULT keyword.

[greptile-apps]

Greptile Summary

This PR implements support for multiple ORDER BY clauses in Centrallix SQL, where later clauses append their sort keys to earlier ones unless a DEFAULT keyword marks a clause as overridable by any subsequent ORDER BY. The change covers parser support, post-processing merge logic, a new bitmask flag, documentation, and seven test cases.

• Parser (mq_internal_SyntaxParse): default is added to the reserved-words list and, when encountered immediately after ORDER BY, sets the new MQ_SF_DEFAULTORDER flag on the clause node.
• Post-processor (mq_internal_PostProcess): A new two-pass merge loop first finds the last ORDER BY, then removes any DEFAULT-flagged clauses that are not last, and finally folds all surviving clauses' children into the first one before expression compilation.
• Tests: Seven CSV-query cases validate single ORDER BY, two-clause append, DEFAULT override, lone DEFAULT, double-DEFAULT, and a three-clause mixed scenario.

Confidence Score: 5/5

• The change is safe to merge. The merge logic is pointer-safe, memory management is correct, and all seven test cases validate the new semantics.
• The core merge algorithm correctly uses pointer identity (not indices) to track the last ORDER BY through removals, children are detached before FreeQS is called so there is no double-free risk, and the ob parameter is cleanly re-derived inside PostProcess rather than relying on the stale parser-side pointer. The only finding is a mildly inaccurate sentence in the documentation.
• The new paragraph in centrallix-doc/SQL.txt has a sentence that inaccurately describes the append behavior for non-DEFAULT ORDER BY clauses and is worth correcting before the docs ship.

Important Files Changed

Filename	Overview
centrallix-doc/SQL.txt	Adds documentation for multi-ORDER BY and DEFAULT keyword; one sentence in the new paragraph inaccurately implies only the last ORDER BY clause is ever appended.
centrallix/multiquery/multiquery.c	Adds ORDER BY merge logic in PostProcess and DEFAULT keyword parsing in SyntaxParse. Memory management is handled correctly (children are moved before FreeQS), pointer-based last_orderby tracking is stable through removals, and the ob parameter is safely re-derived from scratch.
centrallix/include/multiquery.h	Adds MQ_SF_DEFAULTORDER flag (65536) with no conflicts in the existing bitmask sequence.
centrallix/tests/test_sql_orderby_05.to	Seven test cases covering single ORDER BY, multi ORDER BY merge, DEFAULT override, lone DEFAULT, double-DEFAULT, and mixed three-clause scenarios.
centrallix/tests/test_sql_orderby_05.cmp	Expected output for all seven test cases; results are consistent with the described DEFAULT-overwrite semantics.
centrallix-sysdoc/BeeleyCodingStyle.md	Minor prose clarification in the comment-style rule; no functional impact.

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart TD
    A[Parse SQL tokens] --> B{ORDER BY token?}
    B -- No --> C[Other clause handling]
    B -- Yes --> D[Allocate MQ_T_ORDERBYCLAUSE node]
    D --> E{Next token == DEFAULT?}
    E -- Yes --> F[Set MQ_SF_DEFAULTORDER flag]
    E -- No --> G[mlxHoldToken - push back]
    F --> H[Parse ORDER BY items]
    G --> H
    H --> I[Add clause to qs->Children]
    I --> J[Continue parsing...]

    J --> K[mq_internal_PostProcess]
    K --> L[Pass 1: find last ORDER BY node]
    L --> M[Reset ob = NULL, iterate children]
    M --> N{NodeType == ORDERBYCLAUSE?}
    N -- No --> O[Skip, i++]
    N -- Yes --> P{DEFAULT flag AND not last?}
    P -- Yes --> Q[xaRemoveItem + FreeQS, cnt--]
    P -- No --> R{ob == NULL?}
    R -- Yes --> S[ob = subtree, i++]
    R -- No --> T[Move children to ob, xaClear, FreeQS, cnt--]
    S --> U[Compile ORDER BY expressions]
    T --> U
    Q --> M

Loading

_{Reviews (2): Last reviewed commit: "Update default to be a reserved word." | Re-trigger Greptile}

[Lightning11wins]

This PR is ready for human review.

[gbeeley]

A couple of changes needed - thanks!

[gbeeley]

Avoid the goto since this isn't a retry mechanism or an error condition handler, and since this can be easily refactored as an if() block.

[Lightning11wins]

goto removed.

[Lightning11wins]

@gbeeley I think I've resolved your comments.

[gbeeley]

Let's resolve PR 117 first, moving to XArray, and then revisit this one.

[Lightning11wins]

This PR no longer seems to be of trivial size.

[Lightning11wins]

Let's resolve PR 117 first, moving to XArray, and then revisit this one.

The move to XArray for OrderBy in PR #117 has been completed.

[gbeeley]

Looks good - thanks!