Automatically reset buffer when file is read

Problem:

read_csv does not work when multiple files are uploaded

Reason:

It works currently because we are creating a new BytesIO (or StringIO) object each time in deltagenerator. the object gets passed to and gets processed. when another file is uploaded or a rerun happens, we create a new IO object that is starting back at the beginning.

The new implementation does not work because we are using the same buffer and once it’s read the position is not getting reset. need to reset position after read (issues with read_csv 😢)

Several users (here, here, and here) have been stumped by this. We have updated our docs to indicate you should do this.

[ikeshare192]

Excellent…That solved my problems…onward and upward…

[karriebear]

Hi @ikeshare192, sorry to hear about the headache this is causing you. The reason this error is occurring is because of an optimization we made to return the same object on rerun. Previously we were creating a new one on each run which wasn’t ideal. Because we are now returning the same object, the buffer position of the object is persisting.

The solution is to reset the buffer manually like so:

file = st.file_uploader("label")
if file is not None:
	file.seek(0)
        ... 

A workaround would be to use getValue() instead of read() if you are reading the file yourself and not passing this to a library.