Allow customizing row attributes

[SandroHc]

Store the source row attributes, if available, to be used when rendering the table rows.

This introduces a breaking change in rowRender, as the signature of the row argument changed from cellType[] to { cells: cellType[], attributes: object }. I have updated Upgrading.md with the relevant entry for the next major version.

Tested all the demos and did not find regressions. Both DOM and JS init with data are supported; please see demo 25 for how that looks like.

The new tests are finicky when running under Chromium headless. Rows changed order every time I observed them, almost as if the table was being sorted and/or paged - schrodinger's tests. I could not reproduce it on my machine. The only way I managed to make the tests pass consistently was to make sure the tables being tested were small enough as to not have pagination and for the tests to not care about the order of the rows. ¯_(ツ)_/¯

<table>
    <thead>
        <tr>
            <th>Header 1</th>
            <th>Header 2</th>
        </tr>
    </thead>
    <tbody>
        <tr class="my-special-row">
            <td>Cell 1</td>
            <td>Cell 2</td>
        </tr>
    </tbody>
</table>
<script>
    new window.simpleDatatables.DataTable("table")
</script>

[johanneswilm]

@SandroHc accessing and manipulating datatable.data.data will also be working differently with this change, right?

A breaking change is ok as this adds a major useful feature. But let's see if we can bundle it with other major changes so that we don't need to release a version 10 really soon after version 9.

[SandroHc]

You're right, I forgot that those internals are exposed. I have updated the docs and upgrade guide to reflect that.

I have one more change planned, but it can be implemented in a backwards compatible way. It's to allow multiple classes via dt.options.classes. I will try to work on that next week.

[johanneswilm]

I have one more change planned, but it can be implemented in a backwards compatible way. It's to allow multiple classes via dt.options.classes. I will try to work on that next week.

For some of them, it's perfectly fine to use a string containing multiple classes already. In other cases, it is used as a selector so that using something like "class-one class-one class-one" will break the datatable. In those cases, we probably need to choose which one is to be the selector. We could for example pick the first class ("class-one").

[johanneswilm]

@SandroHc Given that this is an example on how to upgrade to version 6.0.x, it seems like this should probably not be changed, right? But maybe this changed example could go under the 8->9 upgrade notes?

[SandroHc]

I've reverted it. Didn't mean to change it, but was in the middle of a fixing spree and didn't notice the file I was on.

[johanneswilm]

Could we add a simple example of what this change means? Just a simple: Instead of X, do Y. So that people who don't read every commit message and just quickly want to upgrade from one to another version quickly figure out what to do.

[SandroHc]

Updated with examples that showcase the changes and the new functionality to customize the row attributes.

I messed up the commit rebase and GitHub is refusing to show the new changes. Please use this link to review them: https://github.com/fiduswriter/simple-datatables/compare/9a6f7f9600af050edbd3aaa945af819d8b40e848..7ccc14c19f765e2a0b954ddb4b743177f7a6813e

[SandroHc]

For some of them, it's perfectly fine to use a string containing multiple classes already. In other cases, it is used as a selector so that using something like "class-one class-one class-one" will break the datatable. In those cases, we probably need to choose which one is to be the selector. We could for example pick the first class ("class-one").

I stumbled across the second case, options.classes.container is used as a selector. I'll try to squeeze a fix this weekend if I manage to get some time.