DataRecs supports 10 connection types across databases, spreadsheets, and BI tools. Extraction correctness for each is proven by a conformance harness that runs a fixed matrix of edge cases (grouping, NULL handling, leading-zero identifiers, Unicode filters, numeric relational operators, and more) against every engine and checks the results agree.
| Source | Identifier | Conformance status |
|---|
| PostgreSQL | postgres | 13/13 conformance cases passing |
| MySQL | mysql | 13/13 conformance cases passing |
| SQL Server | mssql | 13/13 conformance cases passing |
| Oracle | oracle | 13/13 conformance cases passing |
| IBM Db2 | db2 | 13/13 conformance cases passing |
| Google BigQuery | bigquery | 13/13 conformance cases passing |
| Source | Identifier | Conformance status |
|---|
| Google Sheets | google_sheets | Extraction fixtures not yet provisioned in the automated conformance suite — the connector is built and in production use, but this specific edge-case matrix hasn’t been run against it yet. |
| Excel/SharePoint | excel_sharepoint | Same as Google Sheets — fixtures not yet provisioned. |
| Source | Identifier | Conformance status |
|---|
| Tableau | tableau | Not covered by the automated conformance harness (which targets row-level SQL/spreadsheet extraction). |
| Looker | looker | Same as Tableau. |
Each database and BigQuery engine runs the same fixed set of edge cases — grand totals, grouping (including a NULL-valued dimension), string equality/inequality, leading-zero text identifiers (sku = '007'), numeric relational filters (>, >=), IS NULL/IS NOT NULL, a Unicode LIKE filter, and boolean-as-string filtering — against both a golden-CSV oracle and (where available) the live source, and asserts every engine produces identical results. This is what backs the column typing and filtering behaviour described elsewhere in these docs.
