[Haskell-cafe] Projectional editing: Separating a program's AST from its presentation

[Jeffrey Brown]

These ideas were touched on in a previous thread [1].

Our work would be easier if, as data, we separated a program's AST from its
layout.

If, for instance, the order in which a library's functions are presented on
the page were stored as a "projection", separate from the AST that defined
those functions, then one could reorder the functions without obscuring the
history of changes to anything that was "moved".

I assume I am not alone in making a lot of edits aimed solely at allowing
me to read or traverse the code faster? Those within-function changes of
presentation, like across-function changes of order, obscure the record of
functional (as opposed to cosmetic) changes. They don't have to.

Moreover, if projections and the AST were separate data, one could use
contradictory hierarchiies|projections|layouts for the same data. One
projection might group functions by their common purpose (an "is-tree"),
while another grouped them by priority to the reader-rewriter (a "do-tree").

Given current technology, text folding does allow us to impose a hierarchy
on a text document (even a Haskell document! you just need to indent the
code relative to the comments that define the hierarchy).  However, if the
code is stored as text, then only one hierarchy is possible. In particular
one is presently forced to choose between an "is-tree" and a "do-tree".

https://groups.google.com/forum/#!msg/haskell-cafe/d8h8J8VxZxo/BNlKUjWeAgAJ

-- 
Jeffrey Benjamin Brown
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.haskell.org/pipermail/haskell-cafe/attachments/20151024/3a9beb23/attachment.html>