Fast code completion

Thomas Schöps tom.schoeps at gmail.com
Sat Feb 8 12:50:01 GMT 2020


On 2/7/20 4:37 PM, Kevin Funk wrote:
> What did let you go the NIH route and start with CIDE?
Given that I might continue to use an IDE on a daily basis for ~30 more
years, I thought that it would be really helpful if it worked exactly
the way I personally want it to work. Even small things that I am
unhappy with could accumulate a large impact over time, given daily use
over a very long time span. Since it is unlikely that an externally
maintained program would accept all my change requests (since there are
always different opinions on things, which is fine), this would most
likely require me to maintain a set of patches on top of a changing code
base if I used an existing application as base (or fork it, which also
does not help the original application that much). In addition, reading
into existing code can be very difficult and handling pull requests
takes a lot of time. Thus I went the route of starting from scratch,
allowing me to understand the whole codebase, have control over it, and
minimize the dependencies.

That being said, I didn't intend to discuss this project here, I just
had to mention it to say that I actually tried the approach to code
completion that I suggested, and it works well.

On 2/7/20 9:48 PM, Milian Wolff wrote:
> Sounds like this could help a lot indeed, 
> at the cost of increased resource consumption.
Yes, it also slows down the second parse of a document since that will
be a new parse instead of a reparse then. But that might be easy to
avoid if there was a way to clone a translation unit.

Best regards,
Thomas



More information about the KDevelop-devel mailing list