For example, generate Src.Call instead of a chain of Src.App

Build up the canonicalization environment more with dictionaries. Have
the lookup be by name fir...

Centralize data structures in the Source, Valid, and Canonical
categories. This is in contrast t...

Mainly changing imports for the new UTF-8 backing code and generating
the more precise AST struc...

This has a couple benefits:

- More explicit
- Fewer type constraints generated
- No more checki...

This makes things a bit faster weirdly

Now filter out union type constructors that are not direct
dependencies. There are a lot fewer o...

Based on findings of @qbolec in
https://github.com/qbolec/elm-make/commit/81d86e6a745b337a3d0bc1...

Escape characters in String and Char are now limited to the following
set:

\n
\t
\”
\’
\\
\u{00...

Minor grammatical fixes

Port-modules: Typo "thes" fixed

This reduces allocation by a decent amount, from 24% of total to 18% of
total. It does not have ...

The __jsField tokens will be for local renaming within the file. The
__$elmField tokens are for ...

This will produce JS that is easier to mangle. I think it also is a bit
clearer than what I had ...

It converts an int to ASCII by checking if it is in progressively
larger “block sizes” to make s...

Use richer representation of type variables, so FlexVar, FlexSuper,
RigidVar, and RigidSuper are...

I believe that this is a safety precaution, not a necessary step.

Removing this code will set u...

The Type.State now holds a stack of pools. This way when
Type.Solve.generalize moves things to “...

Create NOINLINE values for primitives Types (float, char, string, bool)

Create NOINLINE for unn...

Type.State.flatten now only looks up maxRank once per call, rather than
once for every

The chan...

When adjusting a type alias, I used to crawl the arguments and the
underlying type. All the type...

By switching from a Dict to an Array, I can avoid rebalancing and
allocating quite as much stuff...

Fix typo

Say "1 more argument", not "1 more1 argument".

Fix a few typos in the imports hint

Any level of programmer can be surprised by how `import` works in Elm.
This stuff is always slig...

Fix https://github.com/elm-lang/package.elm-lang.org/issues/185

The idea is that when folks hav...

No dots in names, and names cannot end with dashes.

This is a very roundabout way to fix a big that manifested in elm diff.
When comparing types, so...

Allow more compact representation

Fix #1593 and fix #1594

It is still a function. We are just not providing a name in the error
message.

This way we can decide whether to show ? or ∞ depending on how we
arrived at the error.

Fix #1524
Fix #1552

Not sure if there is a better way to avoid this particular problem. I
hope ...

Preparing to fix a bug!

The OCaml and Haskell implementations both failed in weird ways when
unifying points with themse...

This brings allocation down again. So the entire path of these
optimizations is 914MB => 841MB =...

This changes how `merge` works slightly. Rather than freshly reading
the content to determine th...

This should be cosmetic. No serious changes in how the code is written.

Minimize allocation and be more strict.

In my benchmarks, this reduced allocation from 914MB to...

All symbols are replaced by ASCII letters in a predictable way.