For display_firefly to have Python 3 compatibility, it is necessary that FireflyClient be py3-compatible.

Scope of this work has increased. That is why I changed the SP to 10.

A few outstanding items will be fixed in separate tickets.

See comments on github (https://github.com/lsst/display_firefly/pull/1/files)

I found a number of things while trying these changes out. Some are bugs, some feature requests or merely comments:

1. If you haven't explicitly started an image window with http://$host:$port/firefly/firefly.html;wsch=afw the resulting error isn't very helpful; the exception

   try:

   _fireflyClient = firefly_client.FireflyClient("%s:%d" % (host, port), name)

   _fireflyClient.launch_browser()

   _fireflyClient.add_listener(self.__handleCallbacks)

   except Exception as e:

   raise RuntimeError("Unable to connect to client %s:%d: %s" % (host, port, e))

   propagates out to the prompt.

2. Putting up an image often generates a small image (zoom=1); a small resize of the window causes a resize.
3. It'd be nice to have an option for a horizontal layout (cf. ds9) – displays tend to be wider than high, but the firefly (and default ds9) display is higher than wide so you lose image area.
4. Setting the mask colours in the layers display seems to set their alpha back to 1.0. Also the colour doesn't get set back to the python layer, so if you redisplay the image the colours are reset
5. In the layers display, one layer seems to be the image. Why does it have a colour and a tickbox?
6. a two-finger gesture (on os/x) seems to go to previous/next URLs (I was trying to pan) and I lost my image
7. Pixel coords are off by (0.5, 0.5) – the middle of the bottom left pixel should be (0, 0) (i.e. offset by (1, 1) from the fits standard, which assumes fortran-style 1-indexed arrays)
8. There doesn't seem to be a way to show coordinates allowing for XY0
9. disp.pan(0, 0) is ignored (I think anything below (0.5, 0.5) fails). I'd expect to be able to pan any point to the centre of the display, but I'd certainly expect that it'd clip to a valid coord not silently ignore my request)
10. Killing the firefly display window seems to leave the notebook that's talking to it perfectly happy, but it doesn't respawn the window even if I call the constructor again.
11. The nice inset detail of the image doesn't show the mask planes
12. The inset detail isn't always centred on the cursor. E.g. at the lower left corner, if you put the cursor on the LL pixel, that pixel is shown at the LL of the inset, not the centre
13. Colours for masks appear to be case sensitive (unlike at least most web standards), and illegal colours such as RED are silently ignored
14. display.erase() clears the image, not just the lines/glyphs drawn on it

Thank you Robert Lupton for the comprehensive review and detailed list of issues. Some of these can be addressed in the afwDisplay backend; others will have to be done in Firefly.

Some notes on the ones that could be addressed in the backend:

1. This should be handled better – in the remote case one would not want to launch a remote browser anyway.

14. I thought display.erase() was supposed to clear the image – my mistake. I'll fix it in this ticket.

Suggestions from code review incorporated. A more informative exception message is raised when a browser tab has not be reopened to the correct URL. erase() method now erases only the overlays.

Other issues found in the review will be addressed in new tickets.

Robert Lupton concerning #6: the scrolling works like google maps, which is a click and drag. I have three finger draging set on my Mac (now hidden under: System Preferences... => Accessibility => Mouse & Trackpad => Trackpad Optons... => Enable dragging).

Neither Google maps or Firefly use a "scrolled window" so the OS built in scrolling does not work.

I will make one or two tickets out of the others issues.

Note, #1 and #14 were addressed in the post-review fixes for this ticket, along with the coding comments on the pull request.

The other items were not addressed.

see issue: DM-8409

a two-finger gesture (on os/x) seems to go to previous/next URLs (I was trying to pan) and I lost my image

I think the issue here may be less about "there should be a nice mouse gesture for panning" and more about "it's too easy to lose your place by accidentally invoking 'Back' in the browser".