intersystems-community/iris-agentic-dev/objectscript-tdd

Purpose

Close the feedback loop: write → compile → fix errors → run tests → fix failures → done.

Process Flow

1. Write the class
2. Review — run objectscript-review skill first
3. Compile — run the compile command for this project
4. If compile errors: read the error, identify the line, fix it. Return to step 3.
5. Run tests — run the unit test suite
6. If test failures: read the failure message, fix the code or the test. Return to step 3.
7. Done when: compiles clean + all tests pass

Compile Command

Use the objectscript MCP tool when available (check /mcp):

iris_doc(mode="put", name="MyPackage.MyClass.cls", content="...", compile=true, namespace="USER")

Fallback when MCP is unavailable:

iris session IRIS -U USER "Do $System.OBJ.Load(\"<ClassName>.cls\",\"ck\")"

Key Principle

Never present code to the user that hasn't compiled. If you can't compile (no IRIS access), flag this explicitly:

⚠️ No IRIS connection available — code is unverified. Recommend compiling before use.

A class that compiles clean has not been tested. The compiler does not check that a local is set before it is read, and it cannot check a method name built at runtime. Both of these compile with no error and fail when called:

ClassMethod Total(pA As %Integer, pB As %Integer) As %Integer
{
    Set tSum = pA + pB
    Quit tSun                  // <UNDEFINED> tSun at runtime
}
 
ClassMethod Dyn() As %String
{
    Quit $CLASSMETHOD("MyPackage.MyClass", "Nope")   // <METHOD DOES NOT EXIST> at runtime
}

A literal ..Nope() is caught at compile time, but a dynamic call is not. Call every method you changed at least once, through iris_execute or a test, before you call it done.

Common Compile Errors and Fixes

Related skills

• objectscript-review — run before compiling to catch common mistakes early
• objectscript-unit-test — writing %UnitTest scaffolding
• objectscript-coverage — measuring whether your test suite exercises the code (requires merged toolset)
• objectscript-debugging — when a test fails with a runtime error