[feature request] filter, order and arrange options

[vorpalvorpal]

Thanks for the great project.

I was wondering if you would be interested in implementing any manipulations of the data. View() allows you to reorder and filter rows by a particular column. This would be quite useful to allow users to better explore and edit the data. For example, I want to edit the values of column 1 only where the value of column 2 is greater than x.

Potentially re-ordering columns might also be useful.

If you think this would be useful, the question would be, should the filtering/ordering/arranging be saved or not. It would be easy to filter a column, do some editing and click save without remembering to remove the filter. This would end up with you losing filtered out rows. Equally, however, there would no doubt be use cases where you want the user to be able to filter rows and then return the smaller filtered data frame...

[DillonHammill]

@vorpalvorpal, now that DataEditR uses a modular design (available on the devel branch), it should be possible to add support for such operations. There will be a few things that will need to be figured out:

• These operations will need to be performed independently of rhandsontable as any call to hot_cols() will mess around with other data editing capabilities. Furthermore, as far as I know rhandsontable only supports sorting of numeric columns.
• Based on the above point, this means that we could write dataSort and dataFilter modules to handle these operations. The major question would be what should the UI look like? The reason this is not currently implemented is to maximize the space available to display the data for editing - adding additional UI elements will require more scrolling to view the data.
• To maximize available space, I will need to add arguments to data_edit() to allow control over which modules to include in the application (e.g. logical arguments input, output, edit, sort and filter).
• There will also be additional complexity in getting the modules to appropriately communicate with one another, for example if you edit a filtered dataset but want to update the entire dataset as well.
• Changing column order is supported by rhandsontable but has been deliberately turned off for now as it adds another layer of complexity when converting interactive edits into code (still a work in progress). However, I think it may be possible to add back support for this if I simply compare the output data to the original data and convert this into code.

The key will be in designing an appropriate and minimalist UI to support these operations - I will think about it and see if I can come up with something that will work.

[vorpalvorpal]

That's a good point about the UI. I hadn't really thought about that.

[DillonHammill]

@vorpalvorpal, I just realised that you can actually filter the data by typing in the filtering steps into the data field. For example, if you only wanted to extract/view/edit only the setosa species in iris dataset: This also means that you can do things like head(iris) and tail(iris) to look at the top and bottom rows of the data.

[DillonHammill]

@vorpalvorpal I have added column selection and row filtering modules to the new version of DataEditR that is now available on GitHub.

devtools::install_github("DillonHammill/DataEditR")

[vorpalvorpal]

I just checked it out. Looks awesome! You're a champion.