when a hidden axis is revealed, it is detached from control (reorderable, brushes, etc)

[BigFatDog]

@joshhjacobson has discovered this bug in issue #15
Thank @joshhjacobson for the detailed information

Thanks! This seems to have handled all the console errors, and axis reordering is mostly functional. While testing, I noticed a bug where the right-most axis seems to become disconnected from the canvas. You can still interact with all the other axes by dragging them around and brushing, but the right-most one is unresponsive. If you drag another axis toward it, the polylines will adjust accordingly, but the svg axis itself will remain fixed.

This was produced using the example above where I start with one axis hidden, and then click the button to 'show' it.

[joshhjacobson]

Hi there,

I recently came across an error that might shed some light on this issue. After calling pc.hideAxis(array) I go to update the chart with pc.render().updateAxes(500). This generates the following error:

Which references this line:

So the rightmost axis gets 'detached' because it becomes undefined at some point in this process?

[wraseman]

I believe the issue that @joshhjacobson was running into is because the cars dataset we were using has a column of string data called 'names'. Josh and I were trying to implement some features, like k-means clustering, which perform some operation on the numerical data and the result generates a new column of numerical data. When we inadvertently tried to cluster string and names data, the clustering would work, but the new clustering axis would become unresponsive. After removing the string data, we didn't have an issue.

[joshhjacobson]

@wraseman just to clarify, the string/names data was excluded from the clustering operation. I agree that working with a combination of string and numeric data might be a suspect however. Another important detail here is that all data actually needs to be stored as string values to be compatible with parcoords -- after performing the numeric computations, I would convert all data back to a string format anyway.

One last note: this issue stemmed from a slightly tweaked version of the basic example in this repo (see #15) so I think it's still an issue beyond our specific use case.