9.8 KiB
User's Guide
Basic Concepts
How DataFrames is passed between nodes
DataFrames are transmitted as JSON strings. When a DataFrame enters a node, it is reconstructed by deserializing the JSON string. Upon exiting, it is serialized back into a JSON string.
Pandas's Index object is also processed the same way. For example, Pandas Columns and Pandas Index node return labels in a serialized Index format.
How to connect nodes for DataFrame JSON strings
To connect the DataFrame JSON string output from one Pandas node to another, move the wire toward the top-left of the text field. You may need to hover around the corner to find the correct connection point.
How to add a node
You can right-click, select Add Node, go to Data Analysis, and look for the following nodes:
List of nodes
Data loading, creation, saving
| Node Name | Functionality |
|---|---|
| Pandas Load CSV | Load a CSV file |
| Pandas Load CSV With Encoding | Load a CSV file not encoded in utf-8 |
| Pandas Load CSV With Index | Load a CSV file containing index |
| Pandas Save CSV | Save a DataFrame to a CSV file |
| Pandas Create | Create a DataFrame from CSV text field |
| Pandas Create With Index | Create a DataFrame from CSV text field containing index |
| CDA JSON Create | Create JSON from a string entered in a text field |
Selecting subset of data
| Node Name | Functionality |
|---|---|
| Pandas Columns | Extract column labels |
| Pandas Index | Extract row labels (index) |
| Pandas Select Columns | Select specific columns from a DataFrame |
| Pandas Select Rows | Filter rows based on conditions |
| Pandas Iloc Row Series | Select a row by row integer position |
| Pandas Iloc Rows DataFrame | Select rows by the list of row integer positions |
| Pandas Loc Row Series | Select a row by row label (index) |
| Pandas Loc Cell Str | Select a cell by row label (index) and column label |
| Pandas Head | Extract the first few rows |
Manipulating data
| Node Name | Functionality |
|---|---|
| Pandas Join | Join two DataFrames |
| Pandas Sort | Sort DataFrame by a column |
Displaying data
| Node Name | Functionality |
|---|---|
| Pandas Show DataFrame | Display DataFrame contents |
| Pandas Show Text | Display Text |
| Pandas Summary | Extract DataFrame statistics |
Converting to displayable string
| Node Name | Functionality |
|---|---|
| Pandas Series To String | Convert Series to a displayable string |
| Pandas To String | Convert DataFrame to a displayable string |
Plot, chart
| Node Name | Functionality |
|---|---|
| MPL Bar Chart | Generate a bar chart |
| MPL Line Plot | Generate a line plot |
| MPL Scatter Plot | Generate a scatter plot |
A faster way is to double-click the canvas to open the node search dialog.
The examples directory contains workflows that load data from an example dataset included in this package. These workflows should give you a good idea about how to use these nodes.
Loading data
Use Pandas Load CSV node to load a CSV, or Pandas Load CSV With Index node if the CSV contains an index column. These nodes assume that the file is encoded in utf-8. If the file is not encoded in utf-8, use Pandas Load CSV With Encoding to specify the encoding of the file. This is necessary if the data was collected before utf-8 encoding became popular. For example, you can try "latin_1" for a file that contains accented characters in the US and Western Europe.
These nodes emit the DataFrame in a JSON string. CSV file path is relative to the ComfyUI installation directory unless you specify the absolute file path.
Saving data
Use Pandas Save CSV node to save the DataFrame to a CSV file.
Creating data on the UI
To construct a DataFrame on the UI, use Pandas Create or Pandas Create with Index nodes. These nodes display the multiline text area where you can enter the data in CSV format. If the data contains an index column, use Pandas Create with Index and specify the index column.
Use CDA JSON Create to create JSON from a string entered in a text field. For example, if you want to create a list for a node that requires the list of integers for input, this node is ideal.
Extracting column and row labels
Use Pandas Columns and Pandas Index nodes. As these nodes output JSON-serialized Index object, to display the value, use Pandas Show Text node as shown below:
Selecting columns and rows
Use Pandas Select Columns for selecting columns. For rows, there are multiple options. Use Pandas Select Rows to select rows by a filter condition.
You can use Pandas Loc Row Series to select a row by row label (index). You can use Pandas Iloc Row Series to select a row by row integer position. These nodes returns a JSON-serialized Series object.
You can use Pandas Iloc Rows DataFrame to select rows by the list of row integer positions. This node returns a JSON-serialized DataFrame object.
You need to pass the list of integers to this node. This can be done by using the CDA JSON Create node.
For Pandas Loc Row Series, data type of the labels needs to be specified as shown below:
For Pandas Iloc Row Series, you only need to specify the row integer position:
Note that in Pandas, integer position is an equivalent of an index in other software, while "index" means a unique label assigned to each row. See pandas.DataFrame.iloc.html.
Selecting a cell by row label (index) and column label
Use Pandas Loc Cell Str for selecting a cell by row label (index) and column label. Data type of the labels needs to be specified as shown below:
This node outputs the cell value as a string irrespective of the actual cell data type.
Joining DataFrames
Use Pandas Join. You can select from inner, left, right, outer joins. First, use two Pandas Load CSV nodes to load two CSVs for the left DataFrame and right DataFrame.
Inner join
Specify the common column in the "on column name" field, and set the join method to inner in the Pandas Join node.
Left join
Specify the common column in the "on column name" field, and set the join method to left in the Pandas Join node.
Right join
Specify the common column in the "on column name" field, and set the join method to right in the Pandas Join node.
Outer join
Specify the common column in the "on column name" field, and set the join method to outer in the Pandas Join node.
The workflow file examples/workflows/join.json contains all of the above cases.
Extract the first few rows
Use Pandas Head.
Display text
Use Pandas Show Text node.
Currently this node is using the code taken from ComfyUI-Custom-Scripts's Show Text custom node ( https://github.com/pythongosssss/ComfyUI-Custom-Scripts/blob/main/py/show_text.py). This node has been copied to display text without requiring the user to install the above custom nodes package. However, if you already have ComfyUI-Custom-Scripts, you can also use the Show Text node from the package.
Displaying DataFrame contents
Use Pandas Show DataFrame node. When the number of rows is large, Pandas automatically hides middle rows during string conversion, so the display size should be always manageable.
Extracting DataFrame statistics
Use Pandas Summary to extract statistics then connect the output to Pandas Show Text for display.
Sorting DataFrame by a column
Use Pandas Sort.
Convert DataFrame to a string
Use Pandas To String. This node is to convert the DataFrame to a displayable string, and this truncates the data when the data is large. This can be used to display the data content in a text display node (e.g. Show Text node). However, you can use Show DataFrame nodes to display the DataFrame content, so you do not normally need to use this.
Convert Series to a string
Use Pandas Series To String. This node converts a Pandas Series to a displayable string. You can use this as an input to the Pandas Show Text node.
Displaying charts or plots
Use one of the below:
| Node | Purpose |
|---|---|
| MPL Bar Chart | Generate a bar chart |
| MPL Line Plot | Generate a line plot |
| MPL Scatter Plot | Generate a scatter plot |