Skip to content

feat(workflow-operator): volcano plot significance cut-offs - #9041

Open
eugenegujing wants to merge 9 commits into
apache:mainfrom
eugenegujing:feat/volcano-plot-significance-thresholds
Open

eugenegujing wants to merge 9 commits into
apache:mainfrom
eugenegujing:feat/volcano-plot-significance-thresholds

Conversation

@eugenegujing

@eugenegujing eugenegujing commented Oct 20, 2026 •

Copy link
Copy Markdown
Contributor

What changes were proposed in this PR?

The Volcano Plot draws a plain scatter of effect size vs -log10(p), colored by effect size, which is already the x axis. Nothing shows where the significance bar is or which points clear it. This PR adds three optional properties, all off by default, so saved workflows render as before.

Property Type Effect when set
Significance threshold number, 0 < x <= 1 dashed horizontal line at -log10(threshold)
Effect-size threshold number, > 0 dashed vertical lines at +threshold and -threshold
Color by significance boolean with both thresholds set, points are colored Up / Down / Not significant with a legend instead of the continuous RdBu scale
Before:  px.scatter(color=<effect column>, color_continuous_scale="RdBu")
After:   same by default
         + fig.add_hline(...)                        when a significance threshold is set
         + fig.add_vline(+eff) / fig.add_vline(-eff) when an effect-size threshold is set
         + color=<category column>, 3-color map      when the toggle is on and both thresholds are set

Categories use strict comparisons on the selected p-value column: Up = p < sig and effect > eff, Down = p < sig and effect < -eff, everything else Not significant. The categories are stored in a new column that px colors by. It is named Significance, with underscores appended until no input column has that name, so no input column is overwritten. The legend and the hover call it Significance. The one exception is an effect column that is itself named Significance: the category then shows under the column's own name (Significance_), because px merges two hover lines that share a label. Colors are ColorBrewer RdBu end points #B2182B / #2166AC plus grey #9E9E9E.

Validation: the JSON schema bounds above stop bad values in the property panel. For hand-edited JSON the operator also ignores a threshold it cannot draw (NaN, infinite, <= 0, or a p cut-off > 1) and falls back to the continuous coloring if the toggle is on but a threshold is missing. A cut-off of exactly 1 is allowed and draws its line at y = 0. If the cut-off's -log10 is beyond the data range the line may fall outside the auto-ranged axis; extending the range is left for a follow-up.

Since #8504 merged, the operator also exports a standalone script through generateStandaloneCode(). This PR adds the cut-off lines and the Up / Down / Not significant coloring there too, from the same helpers where possible: the color map and category order are one shared literal, the threshold lines come from the same builder, and only the column names differ (Python literals instead of decoded strings). A spec checks that both paths make the same line and coloring decisions for eight threshold shapes, so the exported figure cannot drift from the engine's.

The label fix (#9037) and the hover column (#9038) have landed and this branch is merged with both. The title and the axis labels apply in both coloring modes; the color bar title only shows with the continuous coloring, since the category coloring has a legend instead of a color bar.

Figures below are the operator's generated Python from main and from this branch, run on the 430-gene example workflow from the issue, with thresholds 0.05 and 1.

Before After (lines only) After (lines + Color by significance)
before-main thresholds-after-lines-only thresholds-after-colored

Any related issues, documentation, discussions?

Closes #9039

PR #8504 (merged) added the generateStandaloneCode() export path; the mirroring described above is the follow-up it needed.

How was this PR tested?

New tests in VolcanoPlotOpDescSpec, written before the source change:

  • defaults: no lines, continuous coloring; each threshold alone emits only its own line(s); both with the toggle off keep the continuous coloring
  • toggle on with both thresholds: strict comparisons, np.select categories written to a column whose name no input column has, color, labels and category_orders keyed by that column, exact color map, no color_continuous_scale
  • an effect column named Significance: neither path assigns table["Significance"], and both leave the Significance label off so the two hover lines stay apart; an effect column named Significance_ keeps the label
  • toggle on with one or no threshold: continuous fallback
  • zero / negative thresholds and a p cut-off above 1 emit no line of their own; Infinity, -Infinity, NaN from JSON never reach the Python
  • a cut-off of exactly 1 draws its line and still colors by category
  • thresholds are emitted as valid Python floats, checked for 1e-300, Double.MIN_VALUE and 1e21 in every site they appear
  • the generated JSON schema carries exclusiveMinimum / maximum on the two thresholds
  • JSON deserialization: numbers, numeric strings, null, blank, absent, and the boolean toggle
  • a workflow JSON saved before this change generates byte-for-byte the same code as a default-constructed operator
  • generateStandaloneCode: defaults draw no line and keep the continuous coloring; the toggle on colors Up / Down / Not significant with the column names as literals; a cross-path test checks that the exported script makes the same cut-off and coloring decisions as the engine for eight threshold shapes (unset, one side only, zero, p cut-off above 1, non-finite, exactly 1)
sbt -Dsbt.server.forcestart=true "WorkflowOperator/testOnly *VolcanoPlotOpDescSpec"
# Tests: succeeded 43, failed 0, canceled 0, ignored 0, pending 1   (the pending one is pre-existing; the count includes the label and hover tests now on main)

sbt -Dsbt.server.forcestart=true "WorkflowOperator/scalafmtAll" "WorkflowOperator/Test/scalafmtAll"
sbt -Dsbt.server.forcestart=true "WorkflowOperator/scalafixAll --check"
# success

Smoke run outside Texera (pytexera stubbed, pandas + plotly 5.24.1) on the 430-gene mock table, 420 rows after the operator drops zero and missing p-values:

Configuration Lines Traces
defaults none 1 continuous trace, 420 points
0.05 / 1, toggle off hline y = 1.301, vlines x = ±1 1 continuous trace
0.05 / 1, toggle on same Up 35, Down 35, Not significant 350
0.05 only, toggle on hline only continuous (fallback)
0.05 / -1, toggle on hline only continuous (fallback)
2 / 1, toggle on vlines only continuous (fallback)
"Infinity" / "NaN" from JSON none continuous, no such token in the code
1e-300 / 1e21, toggle on hline y = 300, vlines x = ±1e21 Not significant 420
1 / 1, toggle on hline y = 0, vlines x = ±1 Up 41, Down 38, Not significant 341

Column-name collisions, run on both generated paths with thresholds 0.05 / 1 and the toggle on (10-row table, 8 rows after the filter): an effect column, a selected hover column or the first column named color; an unused column, the hover column or the p-value column named Significance; Significance and Significance_ both present; an effect column named Significance_; a column name with quotes, a backslash and non-ASCII text. Every case draws Up 2 / Down 1 / Not significant 5 in the three colors with the legend titled Significance, the x values and the hover text equal the input columns, the engine and exported figures are equal as JSON, and the input table is unchanged. With the effect column named Significance the legend reads Significance_ and the hover keeps both the category line and the x line. For an ordinary table the figure is unchanged from the previous revision of this PR.

With thresholds 0.05 / 1, a hover column, and Title, X Label and Y Label all set, both paths draw the configured title and axis labels. With the toggle on they show the Significance legend and no trace uses the color bar; with the toggle off the color bar is titled with the X Label.

The exponent-form dumps pass python -m py_compile. The exported script for the default, lines-only and colored configurations parses with ast.parse.

Was this PR authored or co-authored using generative AI tooling?

Yes. Generated-by: Claude Code (claude-fable-5-1). Co-authored with Claude; the author reviewed the change before submission.

Add optional Significance threshold and Effect-size threshold properties that draw dashed cut-off lines at -log10(threshold) and at plus and minus the effect-size threshold.
Add a Color by significance toggle that, with both thresholds set, colors points Up, Down or Not significant instead of by effect size.
Bound both thresholds in the JSON schema and ignore a non-positive, non-finite or above-one value instead of drawing it.
@github-actions

github-actions Bot commented Oct 20, 2026 •

Copy link
Copy Markdown
Contributor

Automated Reviewer Suggestions

Based on the git blame history of the changed files, we recommend the following reviewers:

  • Contributors with relevant context: @aglinxinyuan
    You can notify them by mentioning @aglinxinyuan in a comment.

@codecov-commenter

codecov-commenter commented Oct 24, 2026 •

Copy link
Copy Markdown

Codecov Report

❌ Patch coverage is 97.56098% with 1 line in your changes missing coverage. Please review.
✅ Project coverage is 91.56%. Comparing base (2e98e44) to head (237888b).
⚠️ Report is 4 commits behind head on main.
✅ All tests successful. No failed tests found.

Files with missing lines Patch % Lines
.../visualization/volcanoPlot/VolcanoPlotOpDesc.scala 97.56% 0 Missing and 1 partial ⚠️
Additional details and impacted files
@@             Coverage Diff              @@
##               main    #9041      +/-   ##
============================================
+ Coverage     91.55%   91.56%   +0.01%     
- Complexity     5112     5139      +27     
============================================
  Files          1291     1291              
  Lines         57359    57398      +39     
  Branches       7502     7502              
============================================
+ Hits          52516    52558      +42     
+ Misses         2933     2932       -1     
+ Partials       1910     1908       -2     
Flag Coverage Δ *Carryforward flag
access-control-service 77.44% <ø> (ø)
agent-service 99.56% <ø> (ø) Carriedforward from 2e98e44
amber 85.32% <97.56%> (+0.04%) ⬆️
computing-unit-managing-service 64.02% <ø> (ø)
config-service 87.50% <ø> (ø)
file-service 82.16% <ø> (ø) Carriedforward from 2e98e44
frontend 96.42% <ø> (ø) Carriedforward from 2e98e44
notebook-migration-service 83.77% <ø> (ø)
pyamber 98.50% <ø> (ø) Carriedforward from 2e98e44
workflow-compiling-service 78.30% <ø> (ø)

*This pull request uses carry forward flags. Click here to find out more.

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:
  • 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

@github-actions

github-actions Bot commented Oct 24, 2026 •

Copy link
Copy Markdown
Contributor

⚠️ Benchmark changes need a look

🟢 2 better · 🔴 3 worse · ⚪ 10 noise (<±5%) · 0 without baseline

Compared against main 2e98e44 benchmarked on this same runner, so the delta is largely free of cross-runner hardware noise. The "7d avg" column still reflects the gh-pages dashboard. Treat <±5% as noise unless repeated.

Dashboard · Run

config throughput MB/s latency max Δ latest / 7d
🔴 bs=10 sw=10 sl=64 412 0.252 22,149/34,293/34,293 us 🟢 -32.3% / 🔴 +156.4%
⚪ bs=100 sw=10 sl=64 863 0.527 115,848/129,898/129,898 us ⚪ within ±5% / 🔴 +50.5%
⚪ bs=1000 sw=10 sl=64 974 0.595 1,024,752/1,095,788/1,095,788 us ⚪ within ±5% / 🔴 +36.3%
Baseline details

Latest main 2e98e44 from same runner

config metric PR latest main 7d avg Δ latest Δ 7d
bs=10 sw=10 sl=64 throughput 412 tuples/sec 441 tuples/sec 1,046 tuples/sec -6.6% -60.6%
bs=10 sw=10 sl=64 MB/s 0.252 MB/s 0.269 MB/s 0.638 MB/s -6.3% -60.5%
bs=10 sw=10 sl=64 p50 22,149 us 20,296 us 10,039 us +9.1% +120.6%
bs=10 sw=10 sl=64 p95 34,293 us 50,629 us 13,374 us -32.3% +156.4%
bs=10 sw=10 sl=64 p99 34,293 us 50,629 us 15,946 us -32.3% +115.1%
bs=100 sw=10 sl=64 throughput 863 tuples/sec 894 tuples/sec 1,408 tuples/sec -3.5% -38.7%
bs=100 sw=10 sl=64 MB/s 0.527 MB/s 0.545 MB/s 0.86 MB/s -3.3% -38.7%
bs=100 sw=10 sl=64 p50 115,848 us 111,387 us 78,599 us +4.0% +47.4%
bs=100 sw=10 sl=64 p95 129,898 us 128,483 us 86,338 us +1.1% +50.5%
bs=100 sw=10 sl=64 p99 129,898 us 128,483 us 108,934 us +1.1% +19.2%
bs=1000 sw=10 sl=64 throughput 974 tuples/sec 982 tuples/sec 1,457 tuples/sec -0.8% -33.2%
bs=1000 sw=10 sl=64 MB/s 0.595 MB/s 0.599 MB/s 0.889 MB/s -0.7% -33.1%
bs=1000 sw=10 sl=64 p50 1,024,752 us 1,011,391 us 766,492 us +1.3% +33.7%
bs=1000 sw=10 sl=64 p95 1,095,788 us 1,111,548 us 804,229 us -1.4% +36.3%
bs=1000 sw=10 sl=64 p99 1,095,788 us 1,111,548 us 850,445 us -1.4% +28.8%
Raw CSV
config_idx,batch_size,schema_width,string_len,num_batches,total_ms,total_tuples,total_bytes,tuples_per_sec,mb_per_sec,lat_p50_us,lat_p95_us,lat_p99_us
0,10,10,64,20,485.06,200,128000,412,0.252,22149.45,34293.14,34293.14
1,100,10,64,20,2318.19,2000,1280000,863,0.527,115847.87,129897.92,129897.92
2,1000,10,64,20,20531.11,20000,12800000,974,0.595,1024752.34,1095787.78,1095787.78

@kz930 kz930 left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM. The schema bounds plus the runtime filter cover both the property panel and hand-edited JSON.

…ignificance-thresholds

# Conflicts:
#	common/workflow-operator/src/main/scala/org/apache/texera/amber/operator/visualization/volcanoPlot/VolcanoPlotOpDesc.scala

@kunwp1 kunwp1 left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM!

…aight to px

Hand the Up / Down / Not significant categories to px as `color` instead of writing them into `table["Significance"]`, in both the engine template and the exported script, and add a regression test. Remove the redundant single-threshold, hold-Doubles, export toggle-off and export saved-workflow tests, stop the engine saved-workflow test at the equality check, and file the two required-field tests back under generatePythonCode.

This branch has not been deployed

No deployments
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Projects

None yet

Development

Successfully merging this pull request may close these issues.

Volcano Plot: optional significance / effect-size thresholds with cut-off lines and up / down / not-significant coloring

4 participants