Dev

[OGR-67]

Description

Closes

closes #49
closes #50

[Copilot]

Pull request overview

This PR adds end-to-end support for AL TableRelation parsing and DBML relationship emission, including schema post-processing to handle missing/partial referenced tables and to keep output consistent when base tables are absent (common with tableextension-only inputs).

Changes:

• Parse TableRelation from AL fields into DBMLColumn.References (with UnknownField sentinel when the referenced field isn’t explicit).
• Extend schema post-processing to remove empty tables, create stub tables for missing referenced targets, and add missing referenced fields to existing partial tables.
• Add/extend tests and fixtures to cover relationship parsing + post-processing behavior; add documentation for parsing and post-processing.

Reviewed changes

Copilot reviewed 8 out of 8 changed files in this pull request and generated 3 comments.

Show a summary per file

File	Description
src/DBMLWriter/SchemaPostProcessor.cs	Adds new post-processing steps for empty tables, stub table creation, and filling missing referenced fields.
src/AL2DBML.Parser/AlParser.cs	Extracts TableRelation into DBMLColumn.References and wires it through table parsing.
src/AL2DBML.Tests/Writer/WriterTests.cs	Adds writer-level tests validating stub creation, empty table removal, and unknown-field resolution.
src/AL2DBML.Tests/Parser/TableParserTests.cs	Adds parser tests for supported/unsupported TableRelation syntaxes (quoted/unquoted/where/conditional).
src/AL2DBML.Tests/Fixtures/Tables/PurchaseHeader.al	Updates fixture to include TableRelation examples used by parser tests.
src/AL2DBML.Tests/Fixtures/Tables/Customer.al	Updates fixture to include TableRelation example used by parser tests.
docs/SCHEMA_POST_PROCESSING.md	New doc describing schema post-processing pipeline and relationship-resolution rules.
docs/AL_PARSING.md	Documents TableRelation parsing and the UnknownField sentinel behavior.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

CreateStubsForMissingReferencedTables marks every distinct explicitly referenced field as IsPrimaryKey = true. If multiple explicit field names reference the same missing table, this produces a composite-PK stub and also prevents ResolveUnknownFieldReferences from resolving UnknownField references to that table (since singlePkByTable only tracks tables with exactly one PK). Consider only marking the stub field as PK when there is exactly one explicit referenced field (otherwise add fields without pk, or choose a single best candidate).

Suggested change

	var explicitFields = group
	.Where(f => !string.IsNullOrEmpty(f.References![1]) && !string.Equals(f.References![1], Unknown, StringComparison.OrdinalIgnoreCase))
	.DistinctBy(f => f.References![1], StringComparer.OrdinalIgnoreCase)
	.Select(f => new DBMLColumn { Name = f.References![1], Type = f.Type, IsPrimaryKey = true })
	.ToList();
	var explicitReferencedFields = group
	.Where(f => !string.IsNullOrEmpty(f.References![1]) && !string.Equals(f.References![1], Unknown, StringComparison.OrdinalIgnoreCase))
	.DistinctBy(f => f.References![1], StringComparer.OrdinalIgnoreCase)
	.ToList();
	var hasSingleExplicitReferencedField = explicitReferencedFields.Count == 1;
	var explicitFields = explicitReferencedFields
	.Select(f => new DBMLColumn
	{
	Name = f.References![1],
	Type = f.Type,
	IsPrimaryKey = hasSingleExplicitReferencedField
	})
	.ToList();

[Copilot]

This link reference appears to be incorrect: ARCHITECTURE.md does not contain a “Schema post-processing” section. Consider updating the reference to SCHEMA_POST_PROCESSING.md (or adding an explicit section/link target in ARCHITECTURE.md).

Suggested change

	When no field is specified, the sentinel value `"UnknownField"` is used. This is resolved at post-processing time (see ARCHITECTURE.md — Schema post-processing).
	When no field is specified, the sentinel value `"UnknownField"` is used. This is resolved at post-processing time (see `SCHEMA_POST_PROCESSING.md`).

[Copilot]

The doc says the stub table contains “that field” as PK, but the current implementation can add multiple distinct explicitly referenced fields for the same missing table (and marks them all as PK). Consider clarifying the behavior for multiple explicit references (single vs multiple stub columns, and PK marking rules) so the documentation matches the code.