Skip to content

Latest commit

 

History

History
169 lines (131 loc) · 5.31 KB

README.md

File metadata and controls

169 lines (131 loc) · 5.31 KB
Raw

Mote

The programmer's assistant. Discharge your burdens with style.

Mote brings the development style found in Agda to Haskell, currently only as a vim plugin.

Requirements

Mote uses pathogen for the vim plugin and assumes you install pathogen bundles to ~/.vim/bundle.

Installation

Pick one of the following.

  1. If you trust me and are okay with programs being installed to /usr/local/bin, just do

    curl https://raw.githubusercontent.com/imeckler/mote/master/scripts/install.sh | bash

    Mote does not set any key bindings for the commands it provides, but a suggested set is given (commented out) here.

  2. If you don't or aren't, do what the script does:

  • First install cabalparse and make sure it's on your path.

  • Install mote via cabal install:

    git clone https://github.com/imeckler/mote.git
cd mote
cabal sandbox init
cabal install mote -j
mv .cabal-sandbox/bin/mote ~/.cabal/bin # or wherever

  • Install the vim plugin (which is in the vim directory of the repo)

  1. Personally, I would do this since it makes updating easy. First install cabalparse and then
mkdir ~/sandboxes # or wherever
cd sandboxes
git clone https://github.com/imeckler/mote.git
cd mote
cabal sandbox init
cabal install mote -j

and then symlink ~/sandboxes/mote/.cabal-sandbox/bin/mote to somewhere on your path. That way, updating is as easy as going to sandboxes/mote, doing a git pull followed by cabal install without having to reinstall dependencies.

Documentation

Mote provides the following vim commands.

  • MoteLoadCurrentFile

    (Re)loads the current file. By default, this is called on every file write. The first load takes a few seconds, but things should be snappy after that.

  • MoteToggleMode

    Toggles whether MoteLoadCurrentFile gets called on every write.

  • MoteNextHole

    Jumps the cursor to the next hole and displays bindings in scope in the hole as well as the goal type. I recommend binding this to option+k.

  • MotePrevHole

    Jumps the cursor to the previous hole and displays bindings in scope in the hole as well as the goal type. I recommend binding this to option+j.

  • Casex

    If x is a variable bound in a pattern match whose right hand side contains the current hole, :Casex x expands the pattern to case further on x. For example, if we have

    f :: [a] -> Int
f xs = _

    then typing :Casex xs results in

    f :: [a] -> Int
f []     = _
f (x:xs) = _

  • CaseOn

    Fills the current hole by casing on some expression. For example, if we have

    f :: [a] -> Int
f xs = _

    then typing :CaseOn xs results in

    f :: [a] -> Int
f xs = case xs of
  []      -> _
  x : xs' -> _

  • MoteRefine

    Partially fill the current hole by providing a function that maps into the goal type. For example, if we have

    f :: [a] -> Int
f xs = _

    then typing :MoteRefine length will result in

    f :: [a] -> Int
f xs = length _

    I recommend binding this to option+r.

  • MoteStart

    Starts a Mote process running in the background. This is called automatically when a Haskell file is loaded and you should never need to call it.

Example session

When you don't know how to complete a program, you type a _ as a placeholder. For example, if we're writing a map function, we would put an underscore _ on the right hand side initially. We'll call our function map' so as not to clash with the Prelude function.

We then start the editing session by calling :MoteLoadCurrentFIle

An underscore expression is called a hole. When programming with Mote, there is a notion of the "current hole": the unfinished expression currently being worked on.

When you want to work on a particular hole, you enter the hole. The command in the Mote vim plugin is :call mote#nextHole(), which jumps your cursor to the next hole.

You'll notice that there is now an @ in the leftmost column, indicating the line of the current hole, and a new window has popped up displaying some information.

The information displayed is the "goal type", the type inferred for the current hole, and the names and types of local variables, which are likely to be useful for filling the hole. (In the future, this window will also display non-local values which are likely to be useful based on the hole type).

We now decide to pattern match on xs, and so we type :Casex xs, which expands a variable bound in a pattern and we jump to the next hole after filling it with _ : _.

Since our goal type is b, and we have a function f that returns a value of type b, we can refine the hole with f. by typing :MoteRefine f.

In general, if you are in a hole of type T, you can refine with an arbitrary expression of type A1 -> ... -> An -> T, which may itself contain holes.

We can then blow through the rest of the definition, alternatingly calling mote#nextHole() and MoteRefine to finally get