The tokenizer can't really solve this case, it needs help from the
parser. ie, tighter coupling...

* Parsing `* <: A` doesn't crash the parser
* `+ <: A` now parses correctly as `(call-pre + (<:...

* Text-based test reduction moved into test_utils
* Remove some obsolete utils
* Rename reduct...

Discovered parsing packages in General

Copy this list from issue #88 to the docs.

Change the parsing of standalone dotted operators such as `.+` to always
parse as `(. +)`, rega...

This is also distinguishable based on the presence of a child `block`
as the only argument, but...

In `Expr` trees, the internal structure of children of `Expr(:.)` is
different within an import...

`JuliaSyntax.parse()` clashes with `Base.parse()` and it's really not clear
whether `parse()` s...

People keep asking me about this 😅

The child layout of try-catch-else-finally is awkward because several of
the subclauses are opt...

Co-authored-by: c42f <[email protected]>

Highlighting of empty ranges needs special handling at the ends of lines
as the source line of ...

* Fix `:kw` conversion in dubious constructs like `f(((a=1)))`
* Fix Expr conversion for parens...

Allow parentheses in import paths and macro names

Some packages use this syntax, even though it's apparently pointless!

Clean up emit_diagnostic stream position handling

`emit_diagnostic(ps, mark, ...)` now behaves like `emit(ps, mark, ...)`
in terms of the source r...

This isn't a complete solution, but it improves the situation somewhat.
(It's unclear whether w...

Introduce a new kind `K"parens"` to represent grouping parentheses with
a tree node of their ow...

Implement a `tokenize()` function which retreives the tokens *after*
parsing.

Going through ...

Various AST cleanups related to removing invisible tokens

Various AST cleanups:

* `struct` and `mutable struct` are distinguished with flags not children...

This allows us to easily distinguish juxtaposition syntax from explicit
multiplication.

`return` can have zero arguments - indicate this with empty children,
rather than with an invisi...

Several syntactic constructs can occur either with or without
parentheses and it can be useful ...

With this change `"str" f` now parses as

(doc (string "str") f)

This better represents...

`highlight()` now always prints whole lines of source code, and can
highlight arbitrary ranges ...

* add a very basic precompile workload

Before:

```
julia> @time @eval collect(JuliaSyntax...

Add basic benchmark script + other rearrangments

Co-authored-by: c42f <[email protected]>

These add a couple more convenience methods useful for TypedSyntax.

The byte range is probably most useful for debugging JuliaSyntax than
general use. So disable t...

Closes #192

Co-authored-by: c42f <[email protected]>

These two files are particularly affected by recent and upcoming
changes (e.g., #193). This add...

The only weird case here is Base's test/syntax.jl which contains some
pretty weird syntax cases...

This was incorrect when not including a finally clause.

- Slightly reword some passages for clarity
- Use punctuation to help reveal sentence structure...

Allow operator-named macros to be used without spaces.

Co-authored-by: c42f <[email protected]>

A fairly big refactor of numeric literal tokenization error cases and a
couple of other tokeniz...

Also add start of more targeted tests for SyntaxNode.

* Line-deletion based fuzzer
* Token-deletion based fuzzer

* Fixes for two crashing bugs fou...

Some fuzz testing tooling to check that the parser doesn't crash on
randomly generated source s...

The reference parser sees `(a; b,)` as a block, but the trailing comma
implies this should be a...

Support for running JuliaSyntax on Julia 1.0

Mostly just compat changes here, adding a bunch of Base utility
functions we use which don't exi...

Fixes for General registry testing

These aren't failures so just ignore them for the purposes of
comparison.

When an interpolation occurs at the beginning of a line without any
preceding whitespace this c...

This syntax is unnecessary but does occur ~10 times in the General
registry so we should suppor...

Here we emit invalid token errors during the token validation pass. This
ensures any invalid to...

Add special cases to explicitly allow a few incompatibilities for cases
where the reference par...

Fix Expr conversion for expressions like `x{a, b; i=j}`

In the reference parser, `[a b\n\n]` parses as `(hcat a b)`, not vcat
with a single row.

Fixes for `is_eventuall_call` so that the following parse correctly

function (::g(x))() e...

Fix cases where parens are used around module name such as

function (A).f()
end

S...

* Compare Expr in detail against reference parser
* Hack for visual expression diffing
* Cache...

This makes the tree more symbolic, by using unique SyntaxNode head types
to refer to core macro...

This "frankenwhere" syntax is used by the SimpleTraits.jl `@traitfn`
macro.

So +0xff is a numeric literal, not `(call-pre + 0xff)`

Several unicode operators were in the wrong precedence category; fix
this:
* Assignment, not c...

jl_strtod_c can return "underflow" even for valid cases such as `5e-324`
where the source strin...