Skip to content
CCL Logo CCL Logo CCL

CRLF Handling in Nested Structures

Normative rule

Section titled “Normative rule”

Implementations that declare the crlf_preserve_literal behavior must handle \r\n line endings correctly during:

  1. Line splitting — Treat \r\n as a single line terminator. Do not let a trailing \r leak into line content.
  2. Indentation detection — Measure leading whitespace only after stripping line terminators. \r must not inflate whitespace counts.
  3. Hierarchy construction — Produce the same nested structure regardless of whether input uses \n or \r\n line endings.

CRLF preservation is a single, indivisible behavior. There is no “flat-only” CRLF mode.

Why

Section titled “Why”

The naive implementation — splitting input on \n and treating lines as opaque — works for flat CCL but silently corrupts nested structures. A trailing \r left on each line:

  • Miscounts indentation (a \r is not whitespace, so depth calculations drift).
  • Breaks equality comparisons on values ("foo\r" ≠ "foo").
  • Shifts comments and values relative to their expected output.

Failures on crlf_preserve_nested_structure and crlf_preserve_comments_and_values (from the test suite) are implementation bugs, not legitimate behavior differences. No new behavior tag or variant is warranted to cover them.

Implementation guidance

Section titled “Implementation guidance”

The canonical fix is to normalize line endings at the earliest possible point in the parsing pipeline:

function parse(text):
    // Before any indentation analysis, split on \r\n OR \n
    lines = split_on_line_terminators(text)  // Strips terminators cleanly
    // Continue with the rest of parse using `lines`

Splitting only on \n and leaving \r attached will pass flat tests and fail nested ones.

Section titled “Related”