Compute rect position and width from samples

[saethlin]

This is my attempt at implementing a solution for #207. Recomputing the percentage position and width every time from the integer sample counts should keep floating-point errors to a minimum. Just that change altered the matched % reported during a search.

But I think the bigger discrepancy was coming from the regex search selecting samples that weren't on screen at all. Since I added the check for hidden objects everything seems quite reasonable.

I'm making this PR because a few days ago I zoomed way in to see the dashmap spinlock loop and the rectangles were hanging over each other. I don't have a reproducer of this on hand to verify, but boy does that seem like floating-point imprecision to me.

[jonhoo]

This is a great change, thank you for taking the time to go through it! I left some comments inline, but in general the approach looks great.

[saethlin]

I just realized that I've broken this code, which attempts to access _orig_x: https://github.com/jonhoo/inferno/pull/210/files#diff-57033a9a1ddd2ffa82ff282a58541526d380285fdebecc47fca185ef5414798dR75-R80

Unfortunately I have no idea what it's for. I can't find any functionality that's missing. Help?

[jonhoo]

Looks like GitHub got confused about your link. I think you're referring to

inferno/src/flamegraph/flamegraph.js

Lines 75 to 80 in 3e249a3

	if (el && el.attributes && el.attributes.y && el.attributes._orig_x) {
	var params = get_params()
	params.x = el.attributes._orig_x.value;
	params.y = el.attributes.y.value;
	history.replaceState(null, null, parse_params(params));
	}

My read of that code is that it checks that the clicked rectangle is actually a sample box, and not just some other random rect on the page. It uses the check that _orig_x is set as a proxy for that. And the preceeding checks are just to avoid errors if it lands on a rect that, say, doesn't have any attributes. You could probably just check for fg:x instead there.

[codecov]

Codecov Report

Merging #210 (675c79c) into master (8fd0289) will decrease coverage by 0.59%.
The diff coverage is 77.14%.

@@            Coverage Diff             @@
##           master     #210      +/-   ##
==========================================
- Coverage   90.08%   89.48%   -0.60%     
==========================================
  Files          17       17              
  Lines        2239     2312      +73     
==========================================
+ Hits         2017     2069      +52     
- Misses        222      243      +21     

Impacted Files	Coverage Δ
src/collapse/mod.rs	52.63% <0.00%> (-14.04%)	⬇️
src/collapse/common.rs	59.66% <50.00%> (-0.61%)	⬇️
src/collapse/guess.rs	69.23% <100.00%> (-8.82%)	⬇️
src/collapse/matcher.rs	100.00% <100.00%> (ø)
src/collapse/perf.rs	95.45% <100.00%> (-1.91%)	⬇️
src/flamegraph/color/mod.rs	83.87% <100.00%> (-0.11%)	⬇️
src/flamegraph/color/palettes.rs	99.31% <100.00%> (-0.23%)	⬇️
src/flamegraph/mod.rs	97.60% <100.00%> (+1.14%)	⬆️
src/flamegraph/svg.rs	97.69% <100.00%> (-0.07%)	⬇️
... and 11 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 43ed780...675c79c. Read the comment docs.

[jonhoo]

Will release this either today or tomorrow, thanks! 👍

[jonhoo]

Released in 0.10.4