Repository navigation
Oracle, and views that know whether they can be written to.
Oracle Database
The fifth engine, and the first to need a change to the generator rather than only a provider and a profile. Oracle returns a generated key through an output bind variable — neither a result set nor a second statement in the batch:
INSERT INTO "CCR"."MEMBER" ( ... ) VALUES ( :p_last_name, ... )
RETURNING "MEMBER_ID" INTO :p_generated_keyThree things behave differently there, and the README covers them properly:
- The Database field is a service name, not a schema —
FREEPDB1on Oracle Free. Getting it wrong produces ORA-50201, whose message blames the syntax of a connect string that is perfectly well-formed. The tool adds its own explanation rather than passing that along unhelped. - The left rail lists schemas, and selecting one filters the relation list instead of reconnecting.
- Identifiers fold to upper case unless they were quoted, and one
NUMBERtype covers booleans, four integer widths and decimals.
Views are read-only when the database says so
A class generated from a view now only gets Add, Update and Delete when writes through it are actually possible. Previously they were always emitted, with a warning in a UI panel that did not travel with the file.
| Engine | How it is determined |
|---|---|
| PostgreSQL | information_schema.views.is_updatable, plus is_trigger_updatable |
| MySQL / MariaDB | information_schema.views.is_updatable |
| Oracle | all_updatable_columns — per column, and accurate |
| SQLite | Read-only unless an INSTEAD OF trigger stands in |
| SQL Server | Its IS_UPDATABLE reports NO for views that accept an UPDATE, so it is not read. A trigger proves updatability; otherwise unknown |
Unknown stays generous — the methods are still generated and the doubt is recorded — because removing capability on a guess would be worse than the problem. A three-state tick on the Generate tab overrules the database either way.
Fixed
- MySQL
DEFAULT CURRENT_TIMESTAMPcolumns were being dropped from every INSERT and UPDATE. MySQL reportsDEFAULT_GENERATEDin a column'sextrafor any default expression, and the check matched the substringgenerated, so everycreated_atandupdated_atcolumn was treated as computed. There was no way to set them and nothing said so. - Selecting an Oracle schema corrupted the connection string, writing the schema into the service-name slot. The failure looked like a syntax error and could not be recovered from the UI.
- A
~in the SQLite database path was not expanded, so~/data/app.dbreached SQLite as a relative path and failed with "unable to open database file" and no path in the message. There is also a Browse… button now. - A newly generated class opens at its first line rather than keeping the previous scroll position.
Verification
The live tests — which compile a generated class with Roslyn and drive real CRUD through it — now run in CI against all five engines, using service containers. On an amd64 runner that includes the genuine SQL Server image.
Downloads
Self-contained single executables. No .NET runtime to install.
| x64 | arm64 | |
|---|---|---|
| Linux | CCReimagined-1.1.0-linux-x64.tar.gz |
CCReimagined-1.1.0-linux-arm64.tar.gz |
| Windows | CCReimagined-1.1.0-win-x64.zip |
CCReimagined-1.1.0-win-arm64.zip |
| macOS | CCReimagined-1.1.0-osx-x64.tar.gz |
CCReimagined-1.1.0-osx-arm64.tar.gz |
Extract and run. On macOS, Gatekeeper needs xattr -d com.apple.quarantine CCReimagined.App — the binaries are not signed or notarised.
Upgrading from 1.0.0
Regenerating a class from a view may remove Add, Update and Delete where the database reports the view as read-only. That is the point of the change, but it is a difference in generated output rather than only an addition — check any view-derived class before replacing it.
Full Changelog: v1.0.0...v1.1.0