Validate a Model

Use validation during authoring, review, and CI.

memo validate [project-directory]

Interactive use

memo validate .

Start with errors, then review warnings. For each finding:

1. locate the element and rule;
2. decide whether the engineering information is missing or modeled with the wrong type or relationship;
3. correct the source;
4. rerun validation.

Machine-readable results

memo validate . --format json --output validation.json
memo validate . --format junit --output validation.xml

Use JSON for custom reporting and JUnit for CI test reports.

External validator checks

A project can select SysIDE as the validator in memo.tools.yaml (or memo.package.yaml), or you can for yourself with memo-architect config set compiler syside (settings are Architect's, not MEMO Tools'). memo validate runs it before MEMO semantic validation and reports its diagnostics as GNU one-liners, carrying SysIDE's own rule codes through unchanged.

toolchain:
  validator: syside
  syside:
    warningsAsErrors: true
    diagnose: all

Or without editing the project, for one run:

memo validate --toolchain.validator syside

An external validator check and a MEMO semantic check answer different questions: the first checks valid and portable language usage; the second checks modelling meaning and traceability.

Diagnostics say which question they answer, in a domain field:

Domain	Meaning
sysml	The selected validator rejected the source.
memo-ingest	The validator accepted it; MEMO could not read all of it.
memo-methodology	Valid SysML, MEMO rule violated.

The distinction matters because MEMO's own grammar covers a subset of SysML. A file SysIDE accepts that MEMO cannot fully read is an ingest gap, reported as memo-ingest — never as a SysML error, and never as a reason to show you nothing.

Suppress and configure by a diagnostic's code, never by matching its message text. Codes are stable across releases; messages are not.

Completeness is a signal, not the objective

Never create unsupported relationships merely to reach 100%. A documented, reviewed gap is safer than a false engineering claim.