[MERGED] Compute patch result at compile time

[PythonNut]

This should move the patch computation to compile time, allowing the byte compiler to optimize the function definitions ahead of time. It also means that el-patch need not be loaded at runtime, unless the user would like to interact with the patches using el-patch's UI.

I'm currently running this, but it may take some work to get it into a mergable state.

[raxod502]

Indeed. As you've probably noticed, the Travis build is failing due to a byte-compilation error. Also, could you make sure there are no lines longer than 80 characters?

Otherwise this looks fantastic. Looking forward to merging.

[raxod502]

Oh yes—the README (new section, I think, for compile-time-only usage) and CHANGELOG (the Unreleased section; see Keep a Changelog) will also need to be updated.

[PythonNut]

@raxod502 for patches to be defined without el-patch at runtime, the user would need to copy the el-patch--patches defvar into their init. Can this be avoided reasonably?

[raxod502]

Can't we make it so that el-patch--patches is initialized if necessary before having a puthash done to it? Then the documentation string will be added if and when el-patch is loaded.

Alternatively, we could autoload el-patch--patches, which is probably a more elegant solution. I'd put in a vote for the latter.

[PythonNut]

Is this what you had in mind?

[raxod502]

Should be fine. I will most likely make some editorial changes so that the style is consistent. But that is easy enough.

I don't think el-patch-feature will work without el-patch-pre-validate-hook also being autoloaded.

[PythonNut]

Hooks are automatically defined if they didn't previously exist, right?

[raxod502]

You're right. I guess there's no need for autoloading the hook.

[raxod502]

Cherry-picked in 4731a62. Please test.