since I spent a bit of time getting used to bokeh, but switched for my actual software to a pure javascript plotting library (dygraphs), I wanted to share with you my experience, which might be interesting for future goals. I’m highly positive about bokeh and wish that it will improve more and more - thank you everyone for the great work.
For me, as a non professional user, the biggest advantage was to be able to write as much as possible in the beautiful (and fast) python language, and reduced the self written, client dependend JS. On the end user side my goal was to have beautiful, modern plots of scientific data.
No, using dygraphs, I didn’t accomplish the first part - I had to write a lot of JS, and even had to do data manipulation on client side, something I really wanted to avoid (if somebody wants to use the code and want’s to work with the data, he has to do this in python and JS now… something I really wanted to avoid). But why did I switch then?
not only the bokeh controls, but also the plots itself don’t work well with touch devices - a must have nowadays in my opinion
the plots often (on some monitors) look “blurry” / not clear/sharp. That’s probably because it’s not svg, but other canvas libraries don’t have this problem as far as I could see. It’s mostly visible on the text, where I would recommend to make the labels/axis stuff as real text (in stylable divs), but also sometimes on the graph lines itself.
I’m missing some functionality like error bands etc. - yes, one can work around this, but should be sooner or later integrated to a scientific tool.
nearly 0 interactivity out of the box (zooming, resizing does not count, that’s obligatory). Have a look at highcharts or amCharts to get an understanding of my point here
So yes, I really understand that this is not “high priority”, because these point’s don’t really touch the core functionality, but basically “nice to have”. And still it hinders people like me to use it productively (especially the touch stuff for websites) and I wanted to share it if there will be the question someday “What can we improve even more?”
Anyways, thank you very much for this great tool and I’ll keep my eyes on it - would love to be able to do all the plotting calculations and stuff on python side one day.
Thanks for the thoughtful commentary. Many of the things you mention are in fact high priorities... just in a very large collection of high priorities. In addition to the people and time constraints that prevent us from working on all the things all the time, in some cases the team simply lacks the necessary expertise. There has been an initial stab at touch support for instance, but it does not work well yet. It would be great to have a new contributor who knows how to make touch UI really shine come aboard and help.
Regarding interactions "out of the box", I'd appreciate more specifics. Things that come standard with Bokeh: pan, zoom, scroll zoom, bounds reset, plot resize, selection tools (point, box, lasso, polygon), hover inspectors (on almost every glyph), easy linked panning, easy linked brushing, easy linked data table, and now all of the sophisticated things that are available with JSCallbacks (a handful of lines of JS required):
I certainly believe there are more things to add, but knowing exactly what you have in mind would he helpful.
Regarding fuzzy text, that's not something I see at all on my day-to-day platform (OS X). There are innumerable combinations of platforms and browsers, and continuous comprehensive testing of all possible combinations is simply not something we can accomplish. To deal with platform specific issues like this one, we very heavily rely on detailed issue reports from users to have any chance to solve them. I encourage you to submit a report to out GH issue tracker with as much information as you can, if you have the time:
Thanks again for taking the effort to write up your thoughts, I hope in the future you can give Bokeh another shot, and that it satisfies your needs.
Bryan
···
On Jul 16, 2015, at 6:41 AM, Clemens Blank <[email protected]> wrote:
Hello everyone,
since I spent a bit of time getting used to bokeh, but switched for my actual software to a pure javascript plotting library (dygraphs), I wanted to share with you my experience, which might be interesting for future goals. I'm highly positive about bokeh and wish that it will improve more and more - thank you everyone for the great work.
For me, as a non professional user, the biggest advantage was to be able to write as much as possible in the beautiful (and fast) python language, and reduced the self written, client dependend JS. On the end user side my goal was to have beautiful, modern plots of scientific data.
No, using dygraphs, I didn't accomplish the first part - I had to write a lot of JS, and even had to do data manipulation on client side, something I really wanted to avoid (if somebody wants to use the code and want's to work with the data, he has to do this in python and JS now.. something I really wanted to avoid). But why did I switch then?
- not only the bokeh controls, but also the plots itself don't work well with touch devices - a must have nowadays in my opinion
- the plots often (on some monitors) look "blurry" / not clear/sharp. That's probably because it's not svg, but other canvas libraries don't have this problem as far as I could see. It's mostly visible on the text, where I would recommend to make the labels/axis stuff as real text (in stylable divs), but also sometimes on the graph lines itself.
- I'm missing some functionality like error bands etc. - yes, one can work around this, but should be sooner or later integrated to a scientific tool.
- nearly 0 interactivity out of the box (zooming, resizing does not count, that's obligatory). Have a look at highcharts or amCharts to get an understanding of my point here
So yes, I really understand that this is not "high priority", because these point's don't really touch the core functionality, but basically "nice to have". And still it hinders people like me to use it productively (especially the touch stuff for websites) and I wanted to share it if there will be the question someday "What can we improve even more?"
Anyways, thank you very much for this great tool and I'll keep my eyes on it - would love to be able to do all the plotting calculations and stuff on python side one day.
