Migrate SND to PostgreSQL

[labkey-adam]

Rationale

We want the SND module to support PostgreSQL

[labkey-adam]

Here's a summary of every translation decision that Claude made when converting the SQL Server SND script to PostgreSQL:

---

SQL Server → PostgreSQL Translation Notes

Type mappings

┌───────────────────────────────────────────────────┬─────────────────────────────────────────────┐
│ SQL Server │ PostgreSQL │
├───────────────────────────────────────────────────┼─────────────────────────────────────────────┤
│ NVARCHAR(n) │ VARCHAR(n) │
├───────────────────────────────────────────────────┼─────────────────────────────────────────────┤
│ NVARCHAR(MAX) │ TEXT │
├───────────────────────────────────────────────────┼─────────────────────────────────────────────┤
│ DATETIME │ TIMESTAMP │
├───────────────────────────────────────────────────┼─────────────────────────────────────────────┤
│ BIT │ BOOLEAN │
├───────────────────────────────────────────────────┼─────────────────────────────────────────────┤
│ BIT NOT NULL DEFAULT 0 │ BOOLEAN NOT NULL DEFAULT FALSE │
├───────────────────────────────────────────────────┼─────────────────────────────────────────────┤
│ BIT NOT NULL DEFAULT 1 │ BOOLEAN NOT NULL DEFAULT TRUE │
├───────────────────────────────────────────────────┼─────────────────────────────────────────────┤
│ UNIQUEIDENTIFIER NOT NULL DEFAULT newid()/NEWID() │ ENTITYID NOT NULL DEFAULT gen_random_uuid() │
├───────────────────────────────────────────────────┼─────────────────────────────────────────────┤
│ UNIQUEIDENTIFIER (nullable) │ ENTITYID │
├───────────────────────────────────────────────────┼─────────────────────────────────────────────┤
│ INTEGER IDENTITY / INT IDENTITY(1,1) │ SERIAL │
├───────────────────────────────────────────────────┼─────────────────────────────────────────────┤
│ date │ DATE │
├───────────────────────────────────────────────────┼─────────────────────────────────────────────┤
│ int │ INTEGER │
├───────────────────────────────────────────────────┼─────────────────────────────────────────────┤
│ LSIDType / LsidType │ LSIDType (same custom domain) │
├───────────────────────────────────────────────────┼─────────────────────────────────────────────┤
│ ENTITYID, USERID │ unchanged (LabKey custom domains) │
└───────────────────────────────────────────────────┴─────────────────────────────────────────────┘

Syntax changes

• Removed all GO batch separators (not valid in PostgreSQL).
• Removed NONCLUSTERED / CLUSTERED index keywords — PostgreSQL has no clustered index concept; all indexes are created
  as plain CREATE INDEX or CREATE UNIQUE INDEX.
• Removed WITH (PAD_INDEX = OFF, STATISTICS_NORECOMPUTE = OFF, ...) index storage options.
• String concatenation + → ||.
• SPACE(n) → REPEAT(' ', n).
• CONVERT(VARCHAR(MAX), expr) → expr::VARCHAR.
• TOP(1) → LIMIT 1.
• EXEC core.fn_dropifexists → SELECT core.fn_dropifexists (not needed here since this is a fresh install script with
  no pre-existing objects).

Functions (fGetSuperPkg, fGetProjectItems)

• SQL Server inline TVF (RETURNS TABLE AS RETURN (...)) → PostgreSQL CREATE OR REPLACE FUNCTION ... RETURNS TABLE
  (...) LANGUAGE sql AS $$ ... $$.
• WITH ... AS CTE → WITH RECURSIVE ... AS (required in PostgreSQL even when the query is recursive).
• Return column types explicitly declared.
• ROW_NUMBER() OVER (...) in the recursive CTE term — PostgreSQL does not permit window functions in the recursive
  member of WITH RECURSIVE. Replaced with COALESCE(sp.SortOrder, 0)::VARCHAR as the path component. This preserves the
  ordering semantics of the original (path segments are zero-padded sort-order values rather than row numbers), and a
  comment in the file explains why.

Function fGetAllSuperPkgs

• The SQL Server cursor loop (CURSOR LOCAL FOR ... FETCH ... WHILE ... INSERT INTO) is replaced with a single CROSS
  JOIN LATERAL against snd.fGetSuperPkg, which is idiomatic PostgreSQL and far more efficient than a procedural loop.
• Corrected the column types in the return table declaration (the original SQL Server script had several wrong types:
  SuperPkgPath INTEGER, Required INTEGER, Active INTEGER, Repeatable INTEGER).

Trigger (ti_after_events)

• SQL Server FOR INSERT AS ... UPDATE ... FROM inserted (statement-level, set-based) → PostgreSQL row-level BEFORE
  INSERT trigger.
• The trigger logic is split into a trigger function (snd.ti_after_events_fn() RETURNS trigger) and a CREATE TRIGGER
  statement (FOR EACH ROW EXECUTE PROCEDURE).
• Using BEFORE INSERT instead of AFTER INSERT means NEW.QcState can be set directly before the row is written,
  avoiding the extra UPDATE that the SQL Server version required.

[labkey-martyp]

Looks good. As discussed we'll loop back on fGetProjectItems and fGetSuperPkg on PG in future testing/PR.

[labkey-martyp]

I'm assuming jdbc will convert this boolean value appropriately to the underlying database

[labkey-adam]

Yes, that's one (of many) benefit of using parameters