Source Model

Pixie's source model supports source-aware diagnostics without forcing every source document to expose a line grid.

Use the model this way:

• SourceDocument is a readable source buffer.
• OriginalSourceDocument owns final diagnostic coordinates.
• StringDocument is a string-backed original document.
• SourceDocumentView and derived documents map generated offsets back to original source.
• SourceSpan represents a contiguous known-or-unknown span.
• SourceRegion represents a possibly sparse set of highlighted characters.
• LineAndColumnPosition uses one-based diagnostic display coordinates.

Offset Rules

Internal offsets are zero-based and half-open:

• Start is the first character offset,
• Length is the number of characters,
• End is one past the final character.

Display coordinates are one-based at API boundaries:

• Line,
• Column.

Reading Text

Open(int offset) should remain useful for large documents and avoid unnecessary large string allocations. GetText(offset, length) may allocate the exact requested text because the caller explicitly asked for that string.

Derived Documents

Generated or preprocessed documents should resolve spans back to original user-authored source. Diagnostics should point at what the user wrote, not at generated intermediate text.

Use SourceSpan for contiguous locations and SourceRegion for sparse highlighting. Preserve unknown spans when APIs can reasonably lack source information.