Closed GoogleCodeExporter closed 9 years ago
UPDATE: Tried to handle the overlapping arrows in the layout-experiment-2
branch; PTAL at that too.
Original comment by s...@seanlip.org
on 23 Mar 2015 at 8:55
Oops, forgot to add link.
https://code.google.com/p/oppia/source/detail?r=f113ceb06a74013e82469c20e365635df5724800&name=layout-experiment-2
(Also, just to be clear, please don't worry about reviewing the code. It's
quite icky! I plan to redo this properly once we figure out what we want the
graphs to look like :P)
Original comment by s...@seanlip.org
on 23 Mar 2015 at 8:56
I think layout-experiment is a *huge* improvement, way to go! For the
overlapping arrows, would it be possible to have the arrows 'stick out' instead
of overlapping with the states below them? I attached a mockup to demonstrate
this.
layout-experiment-2 makes sense to me conceptually but in practice it feels
messy with all the gaps. I think that as an author I wouldn't expect to have
consecutive states be so far apart from each other.
Thanks,
Amit
Original comment by amitdeut...@google.com
on 23 Mar 2015 at 4:44
Attachments:
Thanks for the feedback!
Any thoughts on something in between, like having the nodes indented slightly
(but by less than a full node's width)? I had similar feelings as you did -- I
do like the insight given by layout-experiment-2, but felt a bit uncomfortable
with it and just couldnot put my finger on why.
Original comment by s...@google.com
on 23 Mar 2015 at 4:52
Attached a new screenshot, is that what you meant? Would it be possible to
prototype this in layout-experiment-2 and see how it looks?
Original comment by amitdeut...@google.com
on 23 Mar 2015 at 4:58
Attachments:
I prototyped it in layout-experiment-3 -- sort of. The catch is that it's going
to take some nontrivial work to change where the arrows originate from, and I
don't think I have the time to do this in the near future.
So, I'm going to ask: suppose we could do either (a) what currently exists in
develop, (b) layout-experiment as-is, (c) layout-experiment-2 as-is, or (d)
layout-experiment-3 as-is; which would you prefer? I'll take that as an interim
solution for the time being.
Original comment by s...@seanlip.org
on 24 Mar 2015 at 6:02
Actually, I think layout-experiment-3 is quite nice, that's my top preference.
Well done with this, I think both layout-experiment and layout-experiment-3 are
big improvements over develop.
Original comment by amitdeut...@google.com
on 24 Mar 2015 at 1:14
I like layout-experiment-3! I think the indenting long paths method works quite
well, and it's a significant improvement over develop.
The only strange thing I found is still when a state leads to a bunch of
choices which are linked, and finally leads to another state (this creates the
diagonal structure, which I think does not really express the exploration
structure). I think they should be on the same vertical level, but I'm still
unsure how to implement this. I've looked around and this only happens for the
lazy magician and pitch perfect explorations. For the others
layout-experiment-3 works really well, so that's great.
It may also be good to still have a limit on the max width of the graph.
Original comment by wxy.xi...@gmail.com
on 24 Mar 2015 at 5:36
I think you're right about the "bunch of choices" issue. It would be nice to be
able to detect 'clusters' somehow -- maybe the dominator stuff we were looking
into before might help. But this is probably future work.
Yeah, I am concerned about the max width too -- especially on the history and
stats tabs where the graph does not have its own frame. This could get a bit
complicated, though, because with the indentation we have less space on the
right to stick the extra nodes (and it might also be weird if we abruptly stop
the indentation after some point, e.g. imagine the music exploration). Still,
I'm also thinking that we might want to implicitly encourage explorations
without too much indentation anyway, so maybe it's OK if these edge cases look
a bit weird for now until we find some way to handle them gracefully.
Leaning towards implementing layout-experiment-3 and just putting a hard stop
on indentation at around 3 node widths -- sounds good? If so, I will try and
send a code review request later this week.
Again, thanks a lot for the feedback!
Original comment by s...@google.com
on 24 Mar 2015 at 5:48
Yeah, I'm not sure if stopping the indentation would cause some explorations to
look weird too. Perhaps we can try limiting the indentation and see how it
looks. Why would we want to encourage explorations not to indent too much? For
the stats and history pages, we could also change them to have the graph in a
pan-able window (like the graph modal in the editor).
Original comment by wxy.xi...@gmail.com
on 24 Mar 2015 at 6:01
Hi Xinyu,
I had a go at limiting the indentation. I think it's probably fine; if you
agree, then we probably don't need panning in the other tabs (at least for now).
Re your question, I feel that explorations with lots of indentation are either
intrinsically complicated (which makes them hard to understand and maintain,
which isn't great) or indicative of some common structure like 'choice points'
that we should create an abstraction for (in which case we should just create
it and not have this complexity show up in the graph). I might be wrong,
though, and if you have specific counterexamples then I would be very glad to
hear about them!
Also, I've tidied up the implementation in the better-graph-layout branch.
Please could you (Xinyu) do a proper code review? The link to the commit is
here:
https://code.google.com/p/oppia/source/detail?r=08e91e4c334c930c3a7ec3243b37dd606018a635&name=better-graph-layout
If all looks good I will merge it into develop and it will go out in the next
release.
Thanks!
Original comment by s...@seanlip.org
on 25 Mar 2015 at 7:25
Original comment by s...@seanlip.org
on 20 Apr 2015 at 3:51
Original issue reported on code.google.com by
s...@seanlip.org
on 23 Mar 2015 at 7:35