Skip to content

Add ability to filter out 'very small' changes #1023

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 2 commits into from
Sep 21, 2021
Merged

Add ability to filter out 'very small' changes #1023

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
ae39a5d
Add ability to filter out 'very small' changes
rylev Sep 21, 2021
5a2ec92
Filter our very small changes
rylev Sep 21, 2021
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
1 change: 1 addition & 0 deletions site/src/api.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -201,6 +201,7 @@ pub mod comparison {
pub is_significant: bool,
pub significance_factor: Option<f64>,
pub is_dodgy: bool,
pub magnitude: String,
pub historical_statistics: Option<Vec<f64>>,
pub statistics: (f64, f64),
}
Expand Down
28 changes: 15 additions & 13 deletions site/src/comparison.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -115,6 +115,7 @@ pub async fn handle_compare(
is_dodgy: comparison.is_dodgy(),
is_significant: comparison.is_significant(),
significance_factor: comparison.significance_factor(),
magnitude: comparison.magnitude().display().to_owned(),
historical_statistics: comparison.variance.map(|v| v.data),
statistics: comparison.results,
})
Expand Down Expand Up @@ -149,7 +150,8 @@ async fn populate_report(
}

pub struct ComparisonSummary {
/// Significant comparisons ordered by magnitude from largest to smallest
/// Significant comparisons of magnitude small and above
/// and ordered by magnitude from largest to smallest
comparisons: Vec<TestResultComparison>,
}

Expand All @@ -159,6 +161,7 @@ impl ComparisonSummary {
.statistics
.iter()
.filter(|c| c.is_significant())
.filter(|c| c.magnitude().is_small_or_above())
.cloned()
.collect::<Vec<_>>();
// Skip empty commits, sometimes happens if there's a compiler bug or so.
Expand Down Expand Up @@ -269,27 +272,22 @@ impl ComparisonSummary {
}

pub fn confidence(&self) -> ComparisonConfidence {
let mut num_very_small_changes = 0;
let mut num_small_changes = 0;
let mut num_medium_changes = 0;
for c in self.comparisons.iter() {
match c.magnitude() {
Magnitude::VerySmall => num_very_small_changes += 1,
Magnitude::Small => num_small_changes += 1,
Magnitude::Medium => num_medium_changes += 1,
Magnitude::Large => return ComparisonConfidence::DefinitelyRelevant,
Magnitude::VeryLarge => return ComparisonConfidence::DefinitelyRelevant,
Magnitude::VerySmall => unreachable!(),
}
}

match (
num_very_small_changes,
num_small_changes,
num_medium_changes,
) {
(_, _, m) if m > 1 => ComparisonConfidence::DefinitelyRelevant,
(_, _, m) if m > 0 => ComparisonConfidence::ProbablyRelevant,
(vs, s, _) if (s * 2) + vs > 10 => ComparisonConfidence::ProbablyRelevant,
match (num_small_changes, num_medium_changes) {
(_, m) if m > 1 => ComparisonConfidence::DefinitelyRelevant,
(_, 1) => ComparisonConfidence::ProbablyRelevant,
(s, 0) if s > 10 => ComparisonConfidence::ProbablyRelevant,
_ => ComparisonConfidence::MaybeRelevant,
}
}
Expand Down Expand Up @@ -866,9 +864,9 @@ impl TestResultComparison {
Magnitude::VerySmall
} else if change < threshold * 3.0 {
Magnitude::Small
} else if change < threshold * 10.0 {
} else if change < threshold * 6.0 {
Magnitude::Medium
} else if change < threshold * 25.0 {
} else if change < threshold * 18.0 {
Magnitude::Large
} else {
Magnitude::VeryLarge
Expand Down Expand Up @@ -1002,6 +1000,10 @@ pub enum Magnitude {
}

impl Magnitude {
fn is_small_or_above(&self) -> bool {
*self >= Self::Small
}

fn is_medium_or_above(&self) -> bool {
*self >= Self::Medium
}
Expand Down
140 changes: 83 additions & 57 deletions site/static/compare.html
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -313,7 +313,7 @@ <h2>Comparing <span id="stat-header">{{stat}}</span> between <span id="before">{
<ul id="states-list">
<li>
<label>
<input type="checkbox" id="build-full" v-model="filter.cache.full" />
<input type="checkbox" id="build-full" v-model="filter.scenario.full" />
<span class="cache-label">full</span>
</label>
<div class="tooltip">?
Expand All @@ -324,7 +324,7 @@ <h2>Comparing <span id="stat-header">{{stat}}</span> between <span id="before">{
</li>
<li>
<label>
<input type="checkbox" id="build-incremental-full" v-model="filter.cache.incrFull" />
<input type="checkbox" id="build-incremental-full" v-model="filter.scenario.incrFull" />
<span class="cache-label">incr-full</span>
</label>
<div class="tooltip">?
Expand All @@ -336,7 +336,7 @@ <h2>Comparing <span id="stat-header">{{stat}}</span> between <span id="before">{
<li>
<label>
<input type="checkbox" id="build-incremental-unchanged"
v-model="filter.cache.incrUnchanged" />
v-model="filter.scenario.incrUnchanged" />
<span class="cache-label">incr-unchanged</span>
</label>
<div class="tooltip">?
Expand All @@ -349,7 +349,7 @@ <h2>Comparing <span id="stat-header">{{stat}}</span> between <span id="before">{
<li>
<label>
<input type="checkbox" id="build-incremental-patched"
v-model="filter.cache.incrPatched" />
v-model="filter.scenario.incrPatched" />
<span class="cache-label">incr-patched</span>
</label>
<div class="tooltip">?
Expand All @@ -374,6 +374,19 @@ <h2>Comparing <span id="stat-header">{{stat}}</span> between <span id="before">{
</div>
<input type="checkbox" v-model="filter.showOnlySignificant" style="margin-left: 20px;" />
</div>
<div class="section">
<div class="section-heading"><span>Filter out very small changes</span>
<span class="tooltip">?
<span class="tooltiptext">
Whether to filter out test cases that have a very small magnitude. Magnitude is
calculated both on the absolute magnitude (i.e., how large of a percentage change)
as well as the magnitude of the significance (i.e., by how many time the change was
over the significance threshold).
</span>
</span>
</div>
<input type="checkbox" v-model="filter.filterVerySmall" style="margin-left: 20px;" />
</div>
</div>
</fieldset>
<div v-if="data" id="content" style="margin-top: 15px">
Expand Down Expand Up @@ -430,30 +443,31 @@ <h2>Comparing <span id="stat-header">{{stat}}</span> between <span id="before">{
</tbody>
<template v-for="bench in benches">
<tbody>
<template v-for="run in bench.variants">
<template v-for="testCase in bench.testCases">
<tr>
<th v-if="run.first" v-bind:rowspan="bench.variants.length">{{bench.name}}</th>
<td>{{ run.casename }}</td>
<th v-if="testCase.first" v-bind:rowspan="bench.testCases.length">{{bench.name}}</th>
<td>{{ testCase.scenario }}</td>
<td>
<a v-bind:href="detailedQueryLink(data.a.commit, bench.name, run.casename)">
{{ run.datumA }}
<a v-bind:href="detailedQueryLink(data.a.commit, bench.name, testCase.scenario)">
{{ testCase.datumA }}
</a>
</td>
<td>
<a v-bind:href="detailedQueryLink(data.b.commit, bench.name, run.casename)">
{{ run.datumB }}
<a v-bind:href="detailedQueryLink(data.b.commit, bench.name, testCase.scenario)">
{{ testCase.datumB }}
</a>
</td>
<td>
<a
v-bind:href="percentLink(data.b.commit, data.a.commit, bench.name, run.casename)">
<span v-bind:class="percentClass(run.percent)">
{{ run.percent.toFixed(2) }}%{{run.isDodgy ? "?" : ""}}
v-bind:href="percentLink(data.b.commit, data.a.commit, bench.name, testCase.scenario)">
<span v-bind:class="percentClass(testCase.percent)">
{{ testCase.percent.toFixed(2) }}%{{testCase.isDodgy ? "?" : ""}}
</span>
</a>
</td>
<td>
{{ run.significance_factor ? run.significance_factor.toFixed(2) + "x" :"-" }}
{{ testCase.significanceFactor ? testCase.significanceFactor.toFixed(2) + "x" :"-"
}}
</td>
</tr>
</template>
Expand Down Expand Up @@ -515,7 +529,8 @@ <h2>Comparing <span id="stat-header">{{stat}}</span> between <span id="before">{
filter: {
name: null,
showOnlySignificant: true,
cache: {
filterVerySmall: true,
scenario: {
full: true,
incrFull: true,
incrUnchanged: true,
Expand All @@ -541,51 +556,62 @@ <h2>Comparing <span id="stat-header">{{stat}}</span> between <span id="before">{
let data = this.data;
const filter = this.filter;

function shouldShowBuild(name) {
if (name === "full") {
return filter.cache.full;
} else if (name === "incr-full") {
return filter.cache.incrFull;
} else if (name === "incr-unchanged") {
return filter.cache.incrUnchanged;
} else if (name.startsWith("incr-patched")) {
return filter.cache.incrPatched;
function scenarioFilter(scenario) {
if (scenario === "full") {
return filter.scenario.full;
} else if (scenario === "incr-full") {
return filter.scenario.incrFull;
} else if (scenario === "incr-unchanged") {
return filter.scenario.incrUnchanged;
} else if (scenario.startsWith("incr-patched")) {
return filter.scenario.incrPatched;
} else {
// Unknown, but by default we should show things
return true;
}
}
function toVariants(results) {
let variants = [];

function shouldShowTestCase(testCase) {
let nameFilter = filter.name && filter.name.trim();
nameFilter = !nameFilter || (testCase.benchmark + "-" + testCase.profile).includes(nameFilter);

const significanceFilter = filter.showOnlySignificant ? testCase.isSignificant : true;

const magnitudeFilter = filter.filterVerySmall ? testCase.magnitude != "very small" : true;

return scenarioFilter(testCase.scenario) && significanceFilter && nameFilter && magnitudeFilter;
}

function toTestCases(results) {
let testCases = [];
for (let r of results) {
const scenarioName = r.scenario;
const scenario = r.scenario;
const datumA = r.statistics[0];
const datumB = r.statistics[1];
const isSignificant = r.is_significant;
let percent = (100 * (datumB - datumA) / datumA);
let isDodgy = r.is_dodgy;
if (shouldShowBuild(scenarioName)) {
variants.push({
casename: scenarioName,
datumA,
datumB,
percent,
isDodgy,
significance_factor: r.significance_factor,
isSignificant
});
const significanceFactor = r.significance_factor;
const isDodgy = r.is_dodgy;
let percent = 100 * ((datumB - datumA) / datumA);
let testCase = {
scenario,
datumA,
datumB,
isSignificant,
magnitude: r.magnitude,
significanceFactor,
isDodgy,
percent,
};
if (shouldShowTestCase(testCase)) {
testCases.push(testCase);
}
}

return variants;
return testCases;
}

let benches =
data.comparisons.
filter(n => {
const f = filter.name && filter.name.trim();
return !f || (n.benchmark + "-" + n.profile).includes(f);
}).
reduce((accum, next) => {
const key = next.benchmark + "-" + next.profile;
if (!accum[key]) {
Expand All @@ -598,22 +624,22 @@ <h2>Comparing <span id="stat-header">{{stat}}</span> between <span id="before">{
map(c => {
const name = c[0];
const comparison = c[1];
const variants = toVariants(comparison).filter(v => filter.showOnlySignificant ? v.isSignificant : true);
const pcts = variants.map(field => parseFloat(field.percent));
const testCases = toTestCases(comparison);
const pcts = testCases.map(tc => parseFloat(tc.percent));
const maxPct = Math.max(...pcts).toFixed(1);
const minPct = Math.min(...pcts).toFixed(1);
if (variants.length > 0) {
variants[0].first = true;
if (testCases.length > 0) {
testCases[0].first = true;
}

return {
name,
variants,
testCases,
maxPct,
minPct,
};
}).
filter(b => b.variants.length > 0);
filter(b => b.testCases.length > 0);

const largestChange = a => Math.max(Math.abs(a.minPct), Math.abs(a.maxPct));
// Sort by name first, so that there is a canonical ordering
Expand Down Expand Up @@ -679,7 +705,7 @@ <h2>Comparing <span id="stat-header">{{stat}}</span> between <span id="before">{
},
summary() {
// Create object with each test case that is not filtered out as a key
const filtered = Object.fromEntries(this.benches.flatMap(b => b.variants.map(v => [b.name + "-" + v.casename, true])));
const filtered = Object.fromEntries(this.benches.flatMap(b => b.testCases.map(v => [b.name + "-" + v.scenario, true])));
const newCount = { regressions: 0, improvements: 0, unchanged: 0 }
let result = { all: { ...newCount }, filtered: { ...newCount } }
for (let d of this.data.comparisons) {
Expand Down Expand Up @@ -734,11 +760,11 @@ <h2>Comparing <span id="stat-header">{{stat}}</span> between <span id="before">{
return klass;

},
detailedQueryLink(commit, bench, run) {
return `/detailed-query.html?commit=${commit}&benchmark=${bench}&run_name=${run}`;
detailedQueryLink(commit, bench, testCase) {
return `/detailed-query.html?commit=${commit}&benchmark=${bench}&run_name=${testCase}`;
},
percentLink(commit, baseCommit, bench, run) {
return `/detailed-query.html?commit=${commit}&base_commit=${baseCommit}&benchmark=${bench}&run_name=${run}`;
percentLink(commit, baseCommit, bench, testCase) {
return `/detailed-query.html?commit=${commit}&base_commit=${baseCommit}&benchmark=${bench}&run_name=${testCase}`;
},
commitLink(commit) {
return `https://github.com/rust-lang/rust/commit/${commit}`;
Expand Down Expand Up @@ -823,4 +849,4 @@ <h2>Comparing <span id="stat-header">{{stat}}</span> between <span id="before">{
</script>
</body>

</html>
</html>