| name | objectscript-review |
| description | A concise (<=300 word) pre-write checklist distilling the project's ObjectScript conventions ($$$ macros, Quit-in-try/catch, %OnNew/initvalue, no-underscore names, storage sections untouchable, deploy/test discipline). |
ObjectScript Pre-Write Checklist
Before writing or editing any ObjectScript class, verify:
Macros & naming
- Use
$$$ (triple dollar) for macros, never $$. Multiple $/$$ errors in one file → rewrite the whole file rather than patching piecemeal.
- No underscores in class, method, or parameter names — use camelCase or ALL CAPS instead.
- No
% or _ prefixes on your own properties/parameters (reserved for system classes).
Methods
- Status-returning methods:
Set tSC = $$$OK first line, Quit tSC last line, body wrapped in Try/Catch.
- Inside Try/Catch,
Quit must be ARGUMENTLESS (ERROR #1043 otherwise). Initialize the return variable before the Try block, set it inside, Quit (no args) at the end of Try and Catch, then Quit <var> after the block.
- Abstract methods still need a body returning a sane default (
$$$NULLOREF, "", $$$OK, 0) — never truly empty.
%OnNew(initvalue As %String = "") must call ##super(initvalue) first and check its status; never mark %OnNew Private.
Namespace & REST handlers
- Never
New $NAMESPACE; save/restore explicitly (Set tOrigNS = $NAMESPACE ... Set $NAMESPACE = tOrigNS), restoring BEFORE any catch-block reference to a class from the original namespace.
- Validate inputs before switching to
%SYS.
Errors
- Route error text through
SanitizeError; never embed a caret-global name (^Name) in a message that reaches it — the caret gets silently stripped.
Storage & structure
- NEVER hand-edit a class's Storage section — the compiler owns it.
- Use
/// for doc comments, not // (which does not comment in ObjectScript classes).
- Indent every command at least one space or tab.
Deploy & test
- Create/edit
.cls files on disk first, then iris_doc_load with a glob-prefixed path (a bare path mis-maps the class name), then iris_execute_tests — always compare the returned total against the expected test count before trusting a "0 failures" run; rerun if short.
