Why do mae[[1]] and getWithColData(mae, 1) coexist?

[cvanderaa]

Hi MAE team!

tldr: what were your reasons for having getWithColData(mae, 1) and mae[[1]] as separate functions?

Our QFeatures packages heavily relies on MultiAssayExperiment.

We are rethinking our interface for extracting assays from our objects (see rformassspectrometry/QFeatures#212). We realized that user and developer are often not aware of the getWithColData(mae, 1), but instead use the more naturally feeling mae[[1]] syntax which doesn't manage colData. Conversely, I find myself using getWithColData() most of the time because colData management is much better. See here for an example:

example("MultiAssayExperiment")
colData(mae[[1]]) ## 1.
colData(getWithColData(mae, 1)) ## 2.

In option 1. we only get the colData within the SummarizedExperiment object, wheter 2. provides additional information from the colData in the MultiAssayExperiment object.

I personnally see no reason to prefer behavior in 1. over 2. Therefore, we were thinking reimplementing 1. to rely on getWithColData() and provide as much colData as available. Before doing so, I prefer to hear your opinion and to have your thoughts whether this could be a good or bad idea.

(ping @lgatto)

[LiNk-NY]

Hi @cvanderaa

MultiAssayExperiment stores the experiments as an ExperimentList so that is intrinsically a List type object. MultiAssayExperiment merely passes the [[ method to the inner experiments List. Note that [[ indicates extraction and extraction can be lossless with respect to the rest of the object.
We would prefer to provide the object as-is rather than adding colData without the user's request.
Besides, the inner colData can contain experiment-specific colData and some may use that feature.

The proposed [[ with colData would duplicate the colData columns unnecessarily across all
objects when you run it for multiple experiments (by default). It would also have to resolve any name clashes
and merge with existing colData in the objects.

getWithColData makes it clear that the user wants to duplicate the data from colData at
least for one experiment and merge with existing data in colData in the experiment, if any.

[cvanderaa]

Hello @LiNk-NY,

Thank your prompt and clear answer. This will help us guide our discussion for QFeatures.