fix: parse a keyword list that is a whole inner reference

[jafin]

Fixes #31

The grammar has two alternatives; only one was implemented

FormulaLexer.g4:

fragment INNER_REFERENCE
        : KEYWORD_LIST
        | (KEYWORD_LIST SPACED_COMMA)? COLUMN_RANGE
        ;

TokenParser.ParseIntraTableReference only handled the second. Each keyword block ended with an unconditional SkipComma, which assumes a column range follows the keyword list — so the first alternative walked off the end of the span.

Tracing [[#All]] (length 8):

1. GetArea reads #All, then i += GetLength(All) + 1 → i = 7, and input[7] == ']'. The reference is over.
2. SkipComma(input, 7) is called anyway. Its Debug.Assert(input[i] == ',') is precisely the violated invariant, but it compiles out in Release, so it does i++ and returns 8 == input.Length.
3. GetStructuredName(input, 8, …) reads input[startIdx] and throws IndexOutOfRangeException.

[[#Headers],[#Data]] reaches the same place via the second keyword block.

Every bare-keyword-list form the grammar allows was affected:

Input	Before	After
[[#All]]	IndexOutOfRangeException	All, no columns
[[#Headers]]	IndexOutOfRangeException	Headers, no columns
[[#Data]] / [[#Totals]] / [[#This Row]]	IndexOutOfRangeException	that area, no columns
[[#Headers],[#Data]]	IndexOutOfRangeException	Headers | Data, no columns
[[#Data],[#Totals]]	IndexOutOfRangeException	Data | Totals, no columns
[ [#All] ]	IndexOutOfRangeException	All, no columns

The lexer accepts all of them — AssertTokenType(…, INTRA_TABLE_REFERENCE) passes for each — so this was reachable from any workbook holding one, and surfaced as a runtime exception rather than as ParsingException.

The shorthand [#All] was never affected: it returns early on the KEYWORD branch, before this code.

The fix

Each keyword block now asks whether the inner reference is over — only the SPACED_RBRACKET left — and returns the keyword list with no columns when it is:

private static bool IsEndOfInnerReference(ReadOnlySpan<char> input, int i)
{
    i = SkipWhitespaces(input, i);
    return i >= input.Length || input[i] == ']';
}

A comma is skipped only when one is actually there, so every input that parsed before takes exactly the same path as before. The two ParsingException rejections for the illegal combinations — [[#All],[#Data]] and [[#Headers],[#Totals]], neither of which is a legal KEYWORD_LIST — are untouched and still rejected.

Tests

Two existing tests were already parked on this bug. FormulaConverterToR1C1Tests.StructureReference carried

[InlineData("[[#Headers],[#Data]]", 10, 15, "[[#Headers],[#Data]]", Skip = "Parser fail")]
[InlineData("[[#Data],[#Totals]]", 10, 15, "[[#Data],[#Totals]]", Skip = "Parser fail")]

Both are unskipped here and pass, round-tripping unchanged. Added [[#All]] and [[#Headers]] alongside them, which normalise to the [#All] / [#Headers] shorthand on the way out.

Also added, following each file's existing grammar-comment style:

• IntraTableReferenceTokenTests — all five single keywords, both two-item lists, and the whitespace-padded forms. This test asserts the token type first, so it also pins that the lexer accepts them.
• StructureReferenceRuleTests — the node-level shape, with and without a table name.

[Copilot]

Pull request overview

Fixes a tokenizer crash when parsing structured references whose [[...]] inner-reference consists solely of a keyword list (e.g. [[#All]], [[#Headers],[#Data]]) by correctly treating the keyword list as a complete inner reference with no column range.

Changes:

• Update TokenParser.ParseIntraTableReference to detect end-of-inner-reference after parsing keyword list items and return with firstColumn/lastColumn = null when appropriate.
• Add/extend lexer and rule-level tests to cover single-keyword and two-keyword-list inner references (including whitespace-padded forms).
• Unskip and extend existing R1C1 conversion round-trip tests for previously failing structured references.

Reviewed changes

Copilot reviewed 4 out of 4 changed files in this pull request and generated no comments.

File	Description
src/ClosedXML.Parser/TokenParser.cs	Prevents walking past the end of the intra-table-reference span when the inner reference ends after a keyword list.
src/ClosedXML.Parser.Tests/Rules/StructureReferenceRuleTests.cs	Adds AST-shape coverage for keyword-list-only inner references (with/without table name).
src/ClosedXML.Parser.Tests/Lexers/IntraTableReferenceTokenTests.cs	Adds tokenization coverage for keyword-list-only inner references and whitespace-padded variants.
src/ClosedXML.Parser.Tests/FormulaConverterToR1C1Tests.cs	Unskips previously failing cases and adds round-trip coverage for [[#All]] / [[#Headers]] normalization.

---

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