How to write Good IDE?

[Chris Wright via Digitalmars-d]

On Wed, 01 Feb 2017 15:12:42 +0000, unDEFER wrote:
> And I think how easier may be implement such IDE? From zero, or maybe
> possible to use parts of dmd/gdc?

It's awkward to use dmdfe as a library, mainly because it's not vetted to 
work with the GC. You *can* disable the GC, invoke dmdfe, copy out the 
data you need, and then enable the GC.

This doesn't let you take advantage of the fact that almost all the code 
will be identical between two invocations. It's going to be slow. It won't 
necessarily get you the information you need.

On the plus side, it accepts D pretty well, and someone else gets to 
maintain it for you.

You could write your own. This is a project that might take years on its 
own.

You could fork dmdfe, add your own fields and datastructures for things 
you need that aren't there already, add extra analysis that suits you, and 
change error handling to account for partially completed lines of code.