Issue with Deprecated Functions and Gene Count Handling in liana_aggregate Function

[gomez-dan]

While using the liana_aggregate function in the liana package, I encountered several issues related to deprecated functions and gene filtering. Below is a detailed description of the problems and the associated backtrace for clarity:

1. Gene and Cell Filtering Message:
   When executing the liana_aggregate function, a warning message indicates that 3465 genes and 0 cells were removed due to having no counts. This message suggests that there might be an issue with how data is being handled or filtered before processing. The specific output generated by this warning is not disrupting the workflow but might need investigation to ensure that data is processed correctly.

2. Deprecated invoke() Function:
   The code currently uses the invoke() function, which has been deprecated as of rlang version 0.4.0. The recommendation is to replace it with exec() or inject(). This deprecation warning is critical as it affects the reliability of the package with future versions of rlang.

3. Deprecated progress_estimated() Function in dplyr:
   Additionally, the progress_estimated() function, which was deprecated in dplyr version 1.0.0, is being used. This function appears in the workflow as part of the liana package's operations, and continuing to use this deprecated function could lead to potential issues with future updates of dplyr.

For resolution, it would be beneficial to address these deprecated functions by updating them to their recommended alternatives. Also, investigating the initial gene and cell count removal message for potential improvements in data handling might be necessary.

Backtrace provided indicates the flow of function calls leading to these warnings, which can assist in pinpointing the exact locations in the code that require updates.

[dbdimitrov]

Hi @gomez-dan,

1. Are you saying that your data doesn't have any of those, but liana finds some? It's quite a simple row/colSums that checks if there are any genes/cells in count matrix, if there are some, they get filtered.
2. Yeah, this is because exec in the past had issues with large datasets and passing large objects made LIANA hand for no reason. Hence, I'm using invoke.
3. I will note that one to update it in the future.

Pull requests are also more than welcome :)