# Limitations and diagnostics ## Input validation and encoding The decoder parses JSON and expects the OntoUML JSON structure. It does not run the OntoUML JSON Schema or another structural validation step before traversing the document. Malformed JSON, absent required structures, unexpected types, and unsupported values can therefore raise parser or runtime errors during decoding. UTF-8 is attempted first. A Unicode decoding failure triggers a CP1252 retry and `JSONEncodingFallbackWarning`. The parsed data is otherwise handled identically. ## Information without a normative representation OntoUML Vocabulary 1.1.1 does not directly represent: - order among a Path's point values; - non-empty `propertyAssignments` maps; or - content stored in the legacy diagrammatic `Text.value` field. Path order and property assignments can be recorded as non-normative `rdfs:comment` text. Those comments do not extend the vocabulary and should not be interpreted as a formal graph structure. A non-empty `Text.value` is omitted with `UnsupportedTextValueWarning`; an empty value is omitted silently. ## Invalid and incomplete source values Explicit policies control invalid stereotypes, invalid cardinalities, and unresolved diagrammatic `modelElement` references. A non-error policy can preserve or omit affected information and emits a warning. An `error` policy raises an exception and prevents command-line file output. Invalid `width` or `height` values that are not non-negative integers are logged and omitted. Valid zero values are retained as `xsd:nonNegativeInteger`, in accordance with OntoUML Vocabulary 1.1.1. When initial RDF serialization fails, the file writer retries after normalizing invalid URI references. If writing still fails with an operating-system error, the error is propagated. ## Diagnostic channels The transformation uses two diagnostic channels: - logger messages for progress, supplied defaults, legacy validation and correction activity, and some malformed values; and - Python warnings for encoding fallback, stereotype normalization and policy decisions, cardinality repair, unresolved references, shared batch identity, and non-representable source information. CLI `--silent` and the library's silent execution suppress logger progress and legacy validation messages. They do not suppress Python warnings or exceptions. Applications can capture or filter warnings through Python's `warnings` module. ## Reconstruction qualification Complete-project output retains project and supported diagrammatic resources, but it is not a guarantee of byte-for-byte or semantically lossless source reconstruction. JSON formatting, object-key order, null fields, and unsupported or policy-omitted information are not preserved as source JSON. Non-normative comments can retain selected source context, but consumers need explicit logic to interpret them. See [Policies and configuration](policies.md) for exact consequences and [Transformation overview](transformation.md) for processing order.