With the new builder stuff, it is easier to provide a map from raw
names to canonical names. Tha...

This will migrate to another repo

Elm binaries cache some information, and those caches are always
versioned. So bumping now means...

This makes the data smaller at runtime (25% of old size + less
indirection) but the real point i...

Packages have names, but applications do not.

List all of the publicly exposed values. Also switch more things to
text implementations:

- f...

Based on the reasoning outlined in
https://groups.google.com/d/msg/elm-dev/0AHSnDdkSkQ/E0SVU70JEQAJ

This is based on @justinmanley’s contribution that outlined how to
profile the compiler in a nic...

It was out-dated and I’m hoping we can get a proper benchmarking
library written in Elm at some ...

In some cases, I skipped reversing things back to their original order.
In this case, the origin...

The check for upper case letters A-F had a comparison going the wrong
way. Rewrote to make it mo...

Fix #1532 and fix #1536 by trying to parse normal expressions first.
This way you commit to (.x)...

Just need to make it so a .field can follow a tuple term. The case we
are interested in is tuple...

See https://groups.google.com/d/msg/elm-dev/0AHSnDdkSkQ/E0SVU70JEQAJ
for the reasoning behind th...

Just check for arguments first. This should be more common, so it’s
likely this is a bit faster.

It just cluttered up the error messages, making them more confusing
overall.

Improves https://github.com/elm-lang/error-message-catalog/issues/55

Improvement for
https://github.com/elm-lang/error-message-catalog/issues/96

The context stack now says where each bit of context starts. That means
we can show all the rele...

It now will add the red underline as long as it appears on the last
line of the code snippet. So...

The most important addition is the concept of a stack of “context” so
the error message can say ...

Still does not work when there are multiple paths that all do not work.

This means it’s clear what happens after 26

Big chunks do not work though. Need to create a doc comment parser
using Parse.Primitives.

Still have to make the parse errors work.

Also removes mention of Parsec from codebase, so I be...

Can potentially create the Text.Lazy.Builder directly, though I suspect
the expensive part is ac...

Important fixes include:

- Properly handle record accessors. Dot is no longer an infix
operat...

I figured the order of fields in record types didn’t matter, but it
does for creating record con...

Need to produce a column, that way indentation can be checked multiple
times. Needed when you ha...

Accidentally reversed arguments twice!

Works on integers like 123, hex numbers like 0xfab, and floats like
6.022e23

Switch AST.Declarations to Text and get docComments parsing.

This was rolled into Parse.Expression, where applications, infix ops,
negation, and field access...

Benefits include:

- Almost no backtracking! Only eat whitespace once. The only
backtrack is f...

This design should generate significantly less garbage.

- It looks directly at Word16 without...

Fix #1323

Also fix #1344 and detect duplicate fields in record update syntax

Just parse the whitespace and return () without concat or anything

Still need to do something special for infinite literals like ints and
strings