How do I dynamically change the item type in add_option type=“select”?

[hodakamori]

I think block.add_option can set the type of element to select as follows.
How can I receive this item candidate from node's input?

I would like to create a node that receives a data frame and selects a specific column within that data frame, as in the code example below. However, after the items are first rendered, they seem to remain the same regardless of the node's connection state.

def xy_selecter() -> Block:
    block = Block(name="XY Selecter")

    block.add_input(name="In(df)")
    block.add_output(name="Out(X, df)")
    block.add_output(name="Out(y, df)")

    df = block.get_interface(name="In(df)")
    if df is None:
        items = ["select y"]
    else:
        items = df.columns
    block.add_option(
        name="select y",
        type="select",
        items=items,
    )

    def compute_func(self: Any) -> None:
        df = self.get_interface(name="In(df)")
        y_col = self.get_interface(name="select y")
        y = df[y_col]
        X = df[df.columns != y_col]
        self.set_interface(name="Out(X, df)", value=X)
        self.set_interface(name="Out(y, df)", value=y)

    block.add_compute(compute_func)

    return block

[krish-adi]

hey @hodakamori , you can use the get_option method that is here:

barfi/barfi/block_builder.py

Line 169 in 04ddebb

def get_option(self, name: str):

i am working on the documentation to make this better. there is also a newer version for this project that is being built here: https://github.com/krish-adi/barfi/tree/migrate-to-react-flow

[hodakamori]

I am sorry it has taken so long.
And thank you for your comments.

Sorry, what I want to do is to take the information of the node's input and set it to the items of the option.
In other words, in this code, for example, when the columns of In(df) are [“a”, “b”, “c”], items will be [“a”, “b”, “c”], and when the columns are [“dog”, “cat”, “sheep”], items will be [“dog”, “cat”, “sheep”]. This is what we want.

However, the option items is always [“select y”], which is df is None behavior. From what I can see, it seems that the items in block.add_option are determined when the block is added and then do not change dynamically based on the node's connection state.

We would like to create a node where the user uploads a csv file, reads the csv as a dataframe, and then creates a node that selects a specific column and does some processing.

[krish-adi]

Hey @hodakamori , I understand your request now better. I am afraid that isn't possible at the moment. A block's display options cannot be changed during the runtime. That's possibly a limitation on how streamlit is designed. However, you can use a text input field and display the column names some where in your app such that the user can input the name into the input.

[krish-adi]

Also, @hodakamori , I have updated the library and +v 1.0.0 provides a lot of new interactions and flexibility of saving schemas and the use f compute engine. Plase upgrade if you can.