Drop special support for point spans at the end of files (#31)

Drop special support for point spans at the end of files

This broke point spans at the beginning of lines elsewhere in the
file. Unfortunately, there's no way to disambiguate these currently.
SourceSpanWithContext doesn't have any information about where the
span appears in the context, so for point spans at column 0, it could
be at the beginning of the line or after the trailing newline.

We should eventually add something like
SourceSpanWithContext.indexInContext, but that'll require a breaking
release, since downstream users implement FileSpan as an interface.
6 files changed
tree: 0fe8d0ed90ca51f6c95db27486f53a66d4ce2d59
  1. .gitignore
  2. .test_config
  3. .travis.yml
  4. CHANGELOG.md
  5. LICENSE
  6. README.md
  7. codereview.settings
  8. lib/
  9. pubspec.yaml
  10. test/
README.md

source_span is a library for tracking locations in source code. It's designed to provide a standard representation for source code locations and spans so that disparate packages can easily pass them among one another, and to make it easy to generate human-friendly messages associated with a given piece of code.

The most commonly-used class is the package's namesake, SourceSpan. It represents a span of characters in some source file, and is often attached to an object that has been parsed to indicate where it was parsed from. It provides access to the text of the span via SourceSpan.text and can be used to produce human-friendly messages using SourceSpan.message().

When parsing code from a file, SourceFile is useful. Not only does it provide an efficient means of computing line and column numbers, SourceFile.span() returns special FileSpans that are able to provide more context for their error messages.