Send the row index into a rowFilter function

[rehno-lindeque]

Since the sortable table performs its sorting internally it is not possible to do any client-side paging while maintaining the global sort order for all of the data in the table.

This PR addresses that by adding a rowFilter function which allows for any type of filtering, but specifically includes a sort index to be used for paging.

I slightly prefer a previous attempt where I modified the rowAttrs to take the sort index allowing one to set display: none, which is better for screen readers. This would also allow one to make dynamic changes to the styling based on the sort index of the row.

Unfortunately this required a change in the render function from lazy3 to lazy4 (which seems to be untenable?), or otherwise to completely change the schema being passed into the table (which seemed unwise).

In the absence of a lazy4 function I think that this is the simplest change that makes sense for introducing client-side paging. Please let me know what you think, any comments are welcome!

[ocharles]

I slightly prefer a previous attempt where I modified the rowAttrs to take the sort index allowing one to set display: none, which is better for screen readers.

Could you expand on why that is better for screen readers?

[rehno-lindeque]

I think that a similar argument applies as to the hidden html5 attribute: https://stackoverflow.com/a/6708403/167485. It's not too relevant to the application we're working on, but I thought that someone else might care.

[ocharles]

I'm still misunderstanding exactly why you want display: none over not even having the element. What changes for a screen reader?

[rehno-lindeque]

You'd use a class if you cared about accessibility, rather than inline display: none. I should have been more clear. Something like this:

.visually-hidden {
  display: none;
}

@media print {
  .visually-hidden {
    display: block;
  }
}

@media speech {
  .visually-hidden {
    display: block;
  }
}

Pretty useful for printing pages also.

Anyway, it really just boils down to the old presentation versus semantics argument, that one should not pin down the presentation directly in html (which is for content), but via styling instead.