This allows extracting the recovered tree (containing error nodes)
without using the low-level ...

Dotted call syntax parses into various forms which aren't really
consistent. Especially, Expr i...

Here we unconditionally validate all literals after `parse!()` - ie
processing String and Char ...

This change prevents a crash for various cases of bad syntax involving
macro names. In particul...

The reason to use `__dot__` is to allow `macro __dot__` to be defined in
normal Julia source. B...

Correctly parse dots and other same-precedence chaining after macro
calls:

@A().x
@...

* `.1..` is `.1` followed by `..`
* `0x01..` is `0x01` followed by `..`

Also fixes these cas...

Cater for cases where some parts of the signature are inside and some
outside of the parenthese...

In space sensitive contexts, `~` is parsed as unary or binary depending
on whitespace to the le...

Failure due to overeager error detection of `end` keyword.

* Fix for new `parseall()` API
* Reduce false positives by not reporting cases where reference ...

Here we emit `(a, b; c, d; e, f)` as

(tuple a b (parameters c d) (parameters e f))

whi...

Useful to hold associated trivia (delimiter and var prefix) and indicate
in a clean way that va...

This permits the extra parentheses in things like

function (funcname(some, long, argument...

Here I've replicated the fix from JuliaLang/julia#45024 so that
`const x` (ie, without an assig...

This moves the unnecessary special cases for lists of bindings in `let`
out of the parser and i...

Ensure things like `@S[a,b]` where there's no whitespace between the
`@S` and opening `[` are p...

We now record which precise call syntax was used out of the four
options:
* Prefix calls with ...

A flag is a waste for this; we can just use a different kind for this
horrible edge case.

Here we split off char delimiters in the tokenizer rather than
re-parsing them later during val...

* Simplify `@doc` usage for use in Base bootstrap.
* Separate sysimage precompile into part whi...

Rework JuliaSyntax.parse() public API

`parse()` and `parseall()` were generally pretty inconv...

For Expr there's various cases where = is parsed into a kw head, but
this is inconsistent espec...

This allows REPL completion to work correctly. It works by pattern matching the parse tree, rath...

Some minor fixes

We can use Mmap for mapping IOStream but
* It's not clear that this is necessary
* Removing it...

The reference parser represents `do` syntax with a closure for the second
argument. However, th...

Better "unexpected kw" handling in ternary parsing

* Eliminate the use of Ref to communicate values out of the anonymous
function passed to pars...

Updates README with a link to last lowerer proto (circa 2019), since the branch has been deleted...

This change ensures that strings are always encapsulated within a
`K"string"` internal node of ...

Correctly handle begin as kw in typed comprehension

Add notes about AST differences + minor recovery fix

Non-throwing is_identifier_[start_]char

This allows `a ? b : c` syntax to be clearly distinguished from normal
`if a b else c end` with...

Co-authored-by: Kristoffer Carlsson <[email protected]>

and add a test

Allow newline in range expr after : in parens

Never throw in unescape_julia_string

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

Fix over/underflow in parsing float literals

We probably shoudln't be modifying the memory in a String on the slow
path here, so use a Vector...

It seems we can't use Base.parse for floats because this disallows
underflow. So replicate the l...

Fix early-abort handling in array_parse

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

This won't be valid syntax during lowering, but the parser shouldn't
throw.

Also a little cl...

-2 is numeric literal -2, but ensure .-2 is (call .- 2)

Add check_all_packages script

Fix vect parsing with newline before comma

This allows us to call the flisp parser directly instead of using
Meta.parse which may have bee...

Co-authored-by: Sebastian Pfitzner <[email protected]>

Minor cleanups to core parser hook

Move the hacky code for overriding Core._parse from the sysimage
init module into JuliaSyntax.j...

Reduce invalidation/recompliation for parser usage in sysimage

A fixed world age for the parser prevents the need for recompilation of
the parser due to any us...

Previously, `Meta.parseatom()` would consume trailing whitespace which breaks usages like in `Ba...

If a closing token is found in parse_atom, we still need to emit a
nontrivia token to make the ...

core_parser_hook() should return `nothing` when attempting to parse
statements or atoms, but wh...

Sysimage / precompile fixes

As part of rearranging modules in #40, Base.eof(::Char) was accidentally
defined. It seems clear...