[APM] Always render collapsible library stackframes

[ogupte]

[APM] closes #28919 by always rendering library stackframes as collapsible, even when there's only 1

[jasonrhodes]

@ogupte can you explain the difference between library frames and app frames and stack frames again? I think I learned it at one point but I have no memory of the difference so it's hard to know what changes I'm looking at. Thanks! :D

[jasonrhodes]

From the issue, @formgeist said:

Therefore it's suggested that we always collapse library frames in toggles when there's app frames in the stack trace contents.

Should these all be collapsed by default?

[elasticmachine]

💚 Build Succeeded

• continuous-integration/kibana-ci/pull-request

[ogupte]

@ogupte can you explain the difference between library frames and app frames and stack frames again? I think I learned it at one point but I have no memory of the difference so it's hard to know what changes I'm looking at. Thanks! :D

When Span stackframes are rendered, some of them have the isLibrary flag set on them originating from lines outside of the app code. On render, contiguous app stackframes and library stackframes are grouped together, and displayed differently from each other.

[sorenlouv]

Looks good!
Can you fix the pluralization, so it doesn't say "1 library frames" but "1 library frame"?
Something like this:

{
  "defaultMessage": "{count, plural, one {# library frame} other {# library frames}}",
  "values": {
    "count": stackframes.length
  }
}

[formgeist]

LGTM! @sqren already mentioned the pluralization of frame(s) 👍

[formgeist]

@ogupte Something did occur that stuck with me, so I checked in version 6.6, and it seems like we changed the way we displayed library frames in 6.7 and on. Basically, if there's only library frames available, we choose to uncollapse and show them all since there's no app frames to highlight.

Example of behaviour in 6.6

Current behaviour 6.7+

I guess it must've made it in with the margin tweaks you did previously, just never caught because it's a seemingly rare occurrence (especially when visually testing it). Should we address this in another issue/PR?

[ogupte]

I can address this in this PR, just to clarify: when there are only library frames in the stack trace, we should display them initially expanded.

[formgeist]

Now that I'm looking at it, I actually think we should leave it as we're implementing it now. By always collapsing them, the user will expect this to happen at all time, and it would need their interaction to show them. If we're magically displaying them, I'm not sure they'll be able to tell the toggle is uncollapsed and might mistake them for stack frames. Sorry for the confusion, I was getting carried away by finding that it had changed from the last version.

[elasticmachine]

💔 Build Failed

• continuous-integration/kibana-ci/pull-request

[elasticmachine]

💔 Build Failed

• continuous-integration/kibana-ci/pull-request

[ogupte]

retest

[elasticmachine]

💚 Build Succeeded

• continuous-integration/kibana-ci/pull-request

pluralization addressed in latest PR