Olivier Breuleux

I thought `earlify` would work there, but `gulp-browserify` is deprecated/not maintained, so maybe there's something wrong with it? Anyway, if it works, it should be fine, although you probably shouldn't...

Single quote strings used to compile to `%` documents (`ENode`) (they still do inside `%`, for example if you write: `b % 'Hello {i % world}'`). Then I changed it...

With Symbol, there is no reason to have magic properties that are strings. The documention could be accessed as `fn[Symbol.doc]`, for example. To make it easier, a builtin `help(fn)` or...

I'd rather not parse types from docstrings. I'd rather have something like `[-> String?] fn(String? x, String? y) = ...`, or maybe ``` typeof fn = {String?, String?} -> String?...

You are right, but it doesn't make any difference in practice because trying to set an immutable variable is a compile-time error in EG. Though, I should probably generate `const`...

Ah, yeah, that would be cool! I would have to make some adjustments, though. Since originally I compiled to ES5, optional and rest arguments in EG functions do not compile...

Yeah, I don't know, it's been too long 😬 You can try putting parentheses around `n

The idea that s-expressions are more concise is a bit illusory here, since `(a b c d)` and `(a b,c d)` are the same length. Sure, you'd add a space...

That should work, yes. I haven't paid much heed to the `jurigged.watch` interface since I rarely use it, but I like your suggestion, it is good sugar. I would only...

Using `myia.hypermap.HyperMap` instead of `myia.composite.Elemwise` (using the appropriate leaf function) would work as a quick-fix, but it would bypass Python's `__add__/__radd__` protocol entirely, which may not be the right thing...