fix(FeatherMask): correct negative zero indexing for right/bottom feathering

[alvinttang]

Summary

• Fix FeatherMask node applying feathering to the wrong edge for right and bottom parameters
• When x=0 or y=0 in the feathering loops, -0 in Python equals 0, so output[:, :, -0] indexes the first column instead of the last
• This causes the first column/row to be multiplied twice (by both left+right or top+bottom feathering) while the actual last column/row is never feathered

Before (bug)

for x in range(right):
    feather_rate = (x + 1) / right
    output[:, :, -x] *= feather_rate  # -0 == 0, hits first column!

After (fix)

for x in range(right):
    feather_rate = (x + 1) / right
    output[:, :, -(x + 1)] *= feather_rate  # -1 correctly hits last column

Test plan

• Create a solid white mask (e.g. 256x256)
• Apply FeatherMask with right=50, bottom=50 (left=0, top=0)
• Verify the right edge and bottom edge are feathered (gradient to black)
• Verify the left edge and top edge remain sharp (fully white)
• Previously, the left column and top row would incorrectly get darkened while right/bottom edges stayed white

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro

Run ID: eb58a548-cf8a-4bf1-a2d8-16c8113de2a3

📥 Commits

Reviewing files that changed from the base of the PR and between 730eeec and 0c15a58.

📒 Files selected for processing (1)

• comfy_extras/nodes_mask.py

🚧 Files skipped from review as they are similar to previous changes (1)

• comfy_extras/nodes_mask.py

---

📝 Walkthrough

Walkthrough

The PR updates FeatherMask.execute in comfy_extras/nodes_mask.py to fix off-by-one negative indexing when applying feather gradients: right-edge indexing changed to output[:, :, -(x + 1)] and bottom-edge indexing changed to output[:, -(y + 1), :], adjusting how feather rates are applied to the output mask edges.

🚥 Pre-merge checks | ✅ 4 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 0.00% which is insufficient. The required threshold is 80.00%.	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (4 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title accurately describes the main fix: correcting negative zero indexing for right/bottom feathering in FeatherMask.
Description check	✅ Passed	The description clearly explains the bug, provides before/after code examples, and outlines a comprehensive test plan related to the changeset.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[coderabbitai]

🧹 Nitpick comments (1)

comfy_extras/nodes_mask.py (1)

323-325: Please add a regression test for the right=1 / bottom=1 case.

This fix is easy to accidentally undo because -0 == 0 is non-obvious. A tiny mask assertion that only the last column/row is feathered would lock the behavior in.

Also applies to: 331-333

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In `@comfy_extras/nodes_mask.py` around lines 323 - 325, Add a regression test
that covers the edge cases right=1 and bottom=1 to ensure the feather loops use
negative indexing correctly (i.e., index -1, not -0) and only the final
column/row is modified: create a simple known input mask, call the
mask-feathering function (the code that sets feather_rate and mutates output
using variables right, bottom, and output), and assert that all columns/rows
except the last are unchanged and the last column/row has been multiplied by the
expected feather_rate; add separate assertions for the right loop and the bottom
loop (the blocks using "for x in range(right): ... output[:, :, -(x + 1)] *=
feather_rate" and the analogous bottom code).

🤖 Prompt for all review comments with AI agents

Verify each finding against the current code and only fix it if needed.

Nitpick comments:
In `@comfy_extras/nodes_mask.py`:
- Around line 323-325: Add a regression test that covers the edge cases right=1
and bottom=1 to ensure the feather loops use negative indexing correctly (i.e.,
index -1, not -0) and only the final column/row is modified: create a simple
known input mask, call the mask-feathering function (the code that sets
feather_rate and mutates output using variables right, bottom, and output), and
assert that all columns/rows except the last are unchanged and the last
column/row has been multiplied by the expected feather_rate; add separate
assertions for the right loop and the bottom loop (the blocks using "for x in
range(right): ... output[:, :, -(x + 1)] *= feather_rate" and the analogous
bottom code).

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro

Run ID: 18ea1d38-3b77-46b4-9a12-565ac6656de4

📥 Commits

Reviewing files that changed from the base of the PR and between 3ad36d6 and 730eeec.

📒 Files selected for processing (1)

• comfy_extras/nodes_mask.py

[alexisrolland]

Hello @alvinttang and thank you for your contribution.

Do you have a workflow example to reproduce the issue before the fix please?

Test on master branch (looks fine):

Test on your PR (looks identical):

[alexisrolland]

Ok I see the issue now, there's indeed one pixel feathering on the top and left side of the mask before the fix.

I am merging this, thank you.