[repl] do not show all matches on first tab

[trevnorris]

• Version: All
• Platform: All
• Subsystem: REPL (shipped)

It's an annoyance that the REPL prints out all possibilities the first time hitting the tab. It would be much better if it simply autocompleted as much as possible and waited for a subsequent tab before printing out all matches. Here's an example usages:

> Bu<TAB>.al<TAB>U<TAB>S<TAB>

To produce Buffer.allocUnsafeSlow, but in doing so I get a screen filled with matches that I know I don't want.

Throwing this out there in case anyone wants to take care of it before I do (which I don't plan to in the near future).

[princejwesley]

@mscdex @trevnorris FYI, Its implemented for .editor mode (PR link). If .editor mode is good to go, I'll update PR to support least common prefix based tab completion for default mode.

[Fishrock123]

We should do this with or without .editor support.

[princejwesley]

yes, since editor mode PR has support for LCP Tab completion, I thought of modify it to support for all cases.

[silverwind]

Would very much like this feature. We can probably re-use some code, because this works:

> require('a<TAB>
> require('assert

[trevnorris]

Ooh. How about automatic path searching? e.g. require('./foo<TAB> :)

[silverwind]

How about automatic path searching?

Already kind of works, but it wrongly expands require('./<TAB> to stuff inside node_modules.

Oh, and adding to the wishlist for tab-completion, require('path').<TAB> could suggest all return properties from that function call.

[addaleax]

Oh, and adding to the wishlist for tab-completion, require('path').<TAB> could suggest all return properties from that function call.

That would be awesome, but these kind of things get really tricky, as require()ing a module can have side effects and tab completion never should have any.

[silverwind]

side effects

True, probably not worth the risk and hassle to implement something timeout-based.

[princejwesley]

Incorporated tab completion support in #7275