This avoids using maps and sets with custom equality, which are
moderately inefficient relative to built-in maps and sets on both Node
and the Dart VM. In order to make error messages continue to use the
original identifiers where possible, this updates some of them to rely
more heavily on their associated source spans for context.
We now wrap _withStackFrame() around wider sections of code, including
_loadStylesheet() which handles parse errors, so that the @use/@import
stack frames are available.
In addition to being useful for users of Sass, this will make it
possible for core Sass functions to produce warnings without needing
an explicit reference to the evaluator.
This function didn't expect a number ending in ".0", because they're
normally filtered out by fuzzyAsInt(). However, they can make it
through for very large doubles.
This doesn't have any code changes, but I want to verify that the new
deployment infrastructure works and the only way to do that is to
actually deploy.
The automatic @charset adding is useful in general, but there are
consistently cases where it trips up naïve downstream tools. This
option makes it easier for users to control when it occurs.
This adds a --no-unicode option to disable Unicode span rendering,
decouples repl highlighting from SourceSpan.highlight, and updates
tests to work with the new error highlighting.
It also tightly scopes source spans for statements with children.
Previously, source spans for these nodes extended all the way through
any whitespace that followed the node. This led to messy-looking
multiline span highlights with dart-lang/source_span#25.
Now, StylesheetParser.children doesn't consume trailing whitespace.
Instead, we add a helper method StylesheetParser._withChildren that
parses children, creates the appropriate span, and then consumes the
trailing whitespace.
Previously, evaluator called BinaryOperationExpression.span for each
binary operation it evaluated, which in turn called spanForList() to
create a span covering both child expressions. spanForList() then
called .span for both the left and right child operations *twice*,
leading to exponential behavior.
This is now avoided in three complementary ways:
1. The evaluator avoids eagerly calling AstNode.span, instead keeping
the original AstNode until the span itself needs to be accessed.
This means that a span will only be accessed when an error actually
occurs, and then only one operation's span will be accessed.
2. BinaryOperationExpression.span now iterates through any child
operations before calling their .span methods, so it only performs
O(1) allocations.
3. spanForList() now only calls each AstNode.span once.
This introduces two changes:
1. It changes the epsilon within which two numbers are considered
equal to be an order of magnitude smaller than the numeric
precision. Ruby Sass has always done this, and Dart Sass should
have but did not until now.
2. It parses the component of a number after the decimal point using
double.parse() to avoid accumulating floating point errors.
Always include the error location in JS error messages
I was trying to match Node Sass's behavior by having Error.formatted
property have more detail than Error.message, but our errors rely on
source snippets for context so this just ended up making them
confusing.
This allows us to accurately track the source spans for parenthesized
expressions, which in turn allows us to print accurate error
indications.
Adding a new class for this more accurately represents the structure
of the expression, but it also involves an extra allocation during
parsing and an extra level of nesting during evaluation which could
have a small but real performance impact.
We could alternatively add a package-internal setter for
Expression.span, and update the source span for parenthesized
expressions after they're initially parsed. However, this has its own
downsides: it adds complexity and mutability to the object model; and
many expression classes currently use lazily-generated spans, so
making them settable would require adding extra slots on those
classes.
I decided to go with the extra class because it only adds overhead
when parentheses are actually used in practice, as opposed to adding
overhead to every list/color/etc. The runtime overhead is also likely
to be mitigated if at any point we add a constant-folding step.
When a stylesheet is imported, the parsed stylesheet object is cached
based on its canonical URL. However, the stylesheet.span.sourceUrl was
based on the text of the import that was used to load that stylesheet.
The idea was to make the source URL in stack traces look nicer, but it
meant that relative URLs could be resolved based on the old importer's
URL before being sent to the new importer, which caused bugs.
Now stylesheet.span.sourceUrl is always the canonical URL of the
stylesheet, and thus safe to cache. We then use the import cache to
convert the canonical URL to a human-friendly URL at the point at
which we generate stack traces.
This also deprecates support for relative canonical URLs. The
semantics of these URLs were always unclear, and with the new change
in import internals the old behavior doesn't make much sense. It's
preserved for backwards-compatibility, but deprecated.
