Pandas Ne

Comparing two tables and finding where they differ is a classic data chore - new export vs old export, before vs after a transformation. Pandas Ne is the "not equal" node for that: it compares two DataFrames element by element and returns a DataFrame of True/False marking every cell that doesn't match. It's the natural complement to Pandas Eq in the pack's comparison family.

It's part of ComfyUI-Data-Analysis by Hide Inada (HowToSD), the extension that wraps pandas, Seaborn, and Matplotlib into ComfyUI nodes so data analysis runs in the same graph as image generation. Pandas Ne is a straight passthrough of pandas' DataFrame.ne().

How it works

The node calls a_dataframe.ne(b_dataframe). pandas aligns the two frames on labels and compares positionally. Cells where the values differ → True; equal cells → False; and any comparison involving NaN → True, because in pandas a missing value never equals anything, including another NaN. That last bit is a subtle trap: two tables that are "identical" except both have NaN in the same spot will still light up as not-equal there.

The output is a DATAFRAME of booleans - a difference map. That's the whole product. You can use it directly as a flag table, or chain it into Pandas Boolean Index to pull out the rows where things changed.

The inputs that matter

• a_dataframe - the reference frame.
• b_dataframe - the frame being compared against it.

Both required. Output is a DATAFRAME of booleans with matching shape and labels.

How to install it

Standard pack install:

cd ComfyUI/custom_nodes
git clone https://github.com/HowToSD/ComfyUI-Data-Analysis.git
mv ComfyUI-Data-Analysis data-analysis
pip install -r requirements.txt

or ComfyUI Manager → search "Data analysis" → install → restart → reload. The pack's standard dependencies (pandas, matplotlib, seaborn, scipy, scikit-learn) come with it; no GPU, no model downloads.

Common issues

• Everything reports different - mismatched labels. pandas compares by label alignment, so two tables with different column names or row indices will disagree everywhere. Compare tables that share structure.
• NaN spots flag as different - expected: NaN != NaN in pandas. If missing-vs-missing shouldn't count as a change, clean your data first.
• Comparing against one number - that's Pandas Ne Scalar Float's job; frame-vs-frame is for whole-table diffs.

The mental model: Pandas Eq is "where did they match?", Pandas Ne is "where did they drift?". If you're diffing exports, the second one is usually what you actually want.