Console doesn't display suggestions when you first type _

[cjcenizal]

Kibana version:

6.4.0

Describe the bug:

The first time you type an underscore, the autocomplete doesn't appear. If you follow the underscore with another character, the autocomplete appears. Deleting the character and underscore and subsequently typing an underscore will bring up the autocomplete as expected.

[alisonelizabeth]

@cjcenizal investigated this one a bit.

It looks like there’s logic in place that will not enable autocomplete if the index/start chars have changed. You can see on L865.

So when I type GET logstash-0/, the current token value is:

{type: "url.part", value: "_", index: 4, start: 15}

But when I add the _, it changes to:

{type: "url.slash", value: "/", index: 5, start: 16}

This cause the code starting on L865 to evaluate to true and, therefore, autocomplete is not enabled on L883.

It seems like we would not want this check LAST_EVALUATED_TOKEN.start !== currentToken.start as part of the if statement? I tried to track down the history; looks like it was originally added 5+ years ago, so not much to go off of, other than the code comment // not on the same place or nothing changed, cache and wait for the next time.

Let me know if you have any thoughts/insight into this - thanks!

[jloleysens]

I think what is happening here is the following:

We only want to show autocomplete when a user has typed something and they have not moved to a new token. So stated in terms of the current logic:

1. Are we evaluating the same token we did before with respect to their text position?

// stated as `===` because in the code we halt execution here
lastEvaluatedToken.position.column === currentToken.position.column ||
lastEvaluatedToken.position.lineNumber === currentToken.position.lineNumber

This constraint is kind of tricky because we are saying that if the column OR the line number have changed for the last token then we will not show autocomplete. The result is that there is a lagged response on the autocomplete when we move to a new token; we need the user to provide one more input to trigger autocomplete. This guards against triggering autocomplete when you click around the editor, for instance.

2. Did the value of the token change w.r.t. the current value

lastEvaluatedToken.value !== currentToken.value

If this is true, we know the user provided some kind of input for the current token and we should show autocomplete to help them out.

If both of these conditions are true, then we show autocomplete. Therefore the "lagged" autocomplete will occur for all characters in the example provided (not just _) and for all instances where we change to a new token by typing the first char of the new token.

There are different ways to approach solving this bug, I just wanted to dump some of my thoughts for future context.