5 ms·

Since I have never really liked Matplotlib I would really like to learn Bokeh, but unfortunately its inability to export its visualizations in SVG or similar fo

Since I have never really liked Matplotlib I would really like to learn Bokeh, but unfortunately its inability to export its visualizations in SVG or similar formats makes it kind of useless for me as a scientist wanting to publish my results.

pen2l 11y ago
They're working on it, probably in a month or two you'll be able to do it: https://groups.google.com/a/continuum.io/forum/#!searchin/bokeh/svg/bokeh/biqOu8T23l0/L2yxwHSx3nQJ https://groups.google.com/a/continuum.io/forum/#!searchin/bo... At the current time you can export to png, or maybe export to svg in a roundabout manner, by turning html page into pdf, turning pdf into svg, etc.
- ISL 11y ago
  A png isn't a vector format, so the important property of the SVG (or EPS, PDF) is lost. A vectorized bitmap is still a bitmap.
  - pwang 11y ago
    What resolution are you looking for in print? If you export your raster at 300 dpi, and enable lossless PNG compression, shouldn't it suffice for most print purposes?
    - andrepd 11y ago
      It still has limited resolution and a much higher storage footprint. Plotting charts is the vector graphics use case par excellence. Having svg support is a no brainer.
      - pwang 11y ago
        Of course you're right that we do need to support SVG and vector formats, but it's also not entirely true that "vector >> raster" in all cases. Raster does have limited resolution, but so does your screen or most output devices (yes, even scientific plotters). It's also not necessarily true that it has a higher storage footprint; it depends on the complexity and number of glyphs, the fonts you need to embed for the LaTeX formulae, etc. etc. In large-data cases, raster can be a much more viable visualization transfer medium.
        andrepd 11y ago
        That's true, in practical uses a sufficiently high res raster is indistinguishable from a vector image. Still, it would be nice to have that option.
    - KKKKkkkk1 11y ago
      Often, it's not about print. People sometimes take vector images of plots and squeeze a bunch of them on a single page, with the idea that whoever views the paper online can zoom in if they're interested enough.
- dbyte 11y ago
  Working on SVG? What about EPS instead? For scientific publishing EPS is much better unless your visualisation is pretty much an high entropy one.
  - Rapidwire 11y ago
    Please explain why EPS is better.
    - me2i81 11y ago
      Most imaging for offset printing is done in PostScript, which can render EPS directly, so there's no format conversion. For web publishing that's not the case, and either is probably fine depending on the workflow.
    - Fomite 11y ago
      I've never submitted to a journal that took SVG, nearly everyone I submit to takes EPS.
    - dbyte 11y ago
      Reason is quite simple. EPS is based on PostScript. PDF specification contains a subset of PostScript. Embedding EPS into PDF is trivial and yields high quality results. For a printing publication workflow having EPS files at hand saves a lot of time and problems. SVG is nice too, just needs few more intermediate steps.
amelius 11y ago
But hey, you can save your work in a xkcd style: to_bokeh(fig=None, use_pandas=True, xkcd=False) Taken from: http://bokeh.pydata.org/en/latest/docs/user_guide/compat.html http://bokeh.pydata.org/en/latest/docs/user_guide/compat.htm...
- hk__2 11y ago
  But hey, you can already do that with Matplotlib ;) http://matplotlib.org/xkcd/examples/showcase/xkcd.html http://matplotlib.org/xkcd/examples/showcase/xkcd.html
ISL 11y ago
If you're not wedded to Python, have a gander at Gnuplot. Despite its quirks, it's my favorite plotting tool.
- andrepd 11y ago
  +1 for gnuplot. Amazingly powerful and flexible, and exports to png, pdf, eps, latex, etc. Possibly outdated in some aspects but still unbeatable AFAIK. That being said, this sounds nice for adding interactivity (sliders to vary parameters, that sort of stuff). Useful in presentations and the sort.
- noobermin 11y ago
  I severely dislike Gnuplot, but then again, I am wedded to Python. I'm not sure how versatile it is, at least from the slightly acquainted position I am.
  - bigreddot 11y ago
    FYI it's really dead simple to write new "bokeh" command line tools as Bokeh apps now. We already have several ("bokeh html" and "bokeh json") but we are definitely interested in make a gnuplot-like bokeh command line tool that you can just point at a CSV or log file and get a visualization right out. E.g. something like "bokeh graph foo.log -x time -y connections --tail" would create an automatically updating visualization of a growing log file. Or if anyone is interested in this functionality and wants to help speed along the development, please let us know!
jackgolding 11y ago
I would suggest Plot.ly which is now open source and can export to html (with a lot of JavaScript which generates the SVG)
- ricleal 11y ago
  Plot.ly looks better to me as well. Specially the 3D stuff.
xioxox 11y ago
Try my Veusz plotting app and library - http://home.gna.org/veusz http://home.gna.org/veusz - it has a nice interactive GUI, is python scriptable and extensible via plugins. It's based around Python, Qt and numpy. It can export to PDF, EPS, SVG and bitmap formats.
Ciantic 11y ago
This is just a throw, I've not tried. But Bokeh creates Canvas. There is a Canvas 2 SVG https://github.com/gliffy/canvas2svg https://github.com/gliffy/canvas2svg JavaScript library already made.
bigreddot 11y ago
I definitely empathize. It's a technically challenging problem, there are lots of competing important priorities, and we are a small team that can only humanly accomplish so much work at once. Our vision for Bokeh was to be able to create and share interactive data applications with a minimum of python and little-to-no JS and "web tech" required. We are closing in on that vision, so it is my fervent hope (and intention) that we'll be able to turn to this important priority soon as well.

pen2l 11y ago

They're working on it, probably in a month or two you'll be able to do it: https://groups.google.com/a/continuum.io/forum/#!searchin/bokeh/svg/bokeh/biqOu8T23l0/L2yxwHSx3nQJ https://groups.google.com/a/continuum.io/forum/#!searchin/bo... At the current time you can export to png, or maybe export to svg in a roundabout manner, by turning html page into pdf, turning pdf into svg, etc.

ISL 11y ago
A png isn't a vector format, so the important property of the SVG (or EPS, PDF) is lost. A vectorized bitmap is still a bitmap.
- pwang 11y ago
  What resolution are you looking for in print? If you export your raster at 300 dpi, and enable lossless PNG compression, shouldn't it suffice for most print purposes?
  - andrepd 11y ago
    It still has limited resolution and a much higher storage footprint. Plotting charts is the vector graphics use case par excellence. Having svg support is a no brainer.
    - pwang 11y ago
      Of course you're right that we do need to support SVG and vector formats, but it's also not entirely true that "vector >> raster" in all cases. Raster does have limited resolution, but so does your screen or most output devices (yes, even scientific plotters). It's also not necessarily true that it has a higher storage footprint; it depends on the complexity and number of glyphs, the fonts you need to embed for the LaTeX formulae, etc. etc. In large-data cases, raster can be a much more viable visualization transfer medium.
      - andrepd 11y ago
        That's true, in practical uses a sufficiently high res raster is indistinguishable from a vector image. Still, it would be nice to have that option.
  - KKKKkkkk1 11y ago
    Often, it's not about print. People sometimes take vector images of plots and squeeze a bunch of them on a single page, with the idea that whoever views the paper online can zoom in if they're interested enough.
dbyte 11y ago
Working on SVG? What about EPS instead? For scientific publishing EPS is much better unless your visualisation is pretty much an high entropy one.
- Rapidwire 11y ago
  Please explain why EPS is better.
  - me2i81 11y ago
    Most imaging for offset printing is done in PostScript, which can render EPS directly, so there's no format conversion. For web publishing that's not the case, and either is probably fine depending on the workflow.
  - Fomite 11y ago
    I've never submitted to a journal that took SVG, nearly everyone I submit to takes EPS.
  - dbyte 11y ago
    Reason is quite simple. EPS is based on PostScript. PDF specification contains a subset of PostScript. Embedding EPS into PDF is trivial and yields high quality results. For a printing publication workflow having EPS files at hand saves a lot of time and problems. SVG is nice too, just needs few more intermediate steps.

amelius 11y ago

But hey, you can save your work in a xkcd style: to_bokeh(fig=None, use_pandas=True, xkcd=False) Taken from: http://bokeh.pydata.org/en/latest/docs/user_guide/compat.html http://bokeh.pydata.org/en/latest/docs/user_guide/compat.htm...

hk__2 11y ago
But hey, you can already do that with Matplotlib ;) http://matplotlib.org/xkcd/examples/showcase/xkcd.html http://matplotlib.org/xkcd/examples/showcase/xkcd.html

ISL 11y ago

If you're not wedded to Python, have a gander at Gnuplot. Despite its quirks, it's my favorite plotting tool.

andrepd 11y ago
+1 for gnuplot. Amazingly powerful and flexible, and exports to png, pdf, eps, latex, etc. Possibly outdated in some aspects but still unbeatable AFAIK. That being said, this sounds nice for adding interactivity (sliders to vary parameters, that sort of stuff). Useful in presentations and the sort.
noobermin 11y ago
I severely dislike Gnuplot, but then again, I am wedded to Python. I'm not sure how versatile it is, at least from the slightly acquainted position I am.
- bigreddot 11y ago
  FYI it's really dead simple to write new "bokeh" command line tools as Bokeh apps now. We already have several ("bokeh html" and "bokeh json") but we are definitely interested in make a gnuplot-like bokeh command line tool that you can just point at a CSV or log file and get a visualization right out. E.g. something like "bokeh graph foo.log -x time -y connections --tail" would create an automatically updating visualization of a growing log file. Or if anyone is interested in this functionality and wants to help speed along the development, please let us know!

jackgolding 11y ago

I would suggest Plot.ly which is now open source and can export to html (with a lot of JavaScript which generates the SVG)

ricleal 11y ago
Plot.ly looks better to me as well. Specially the 3D stuff.

xioxox 11y ago

Try my Veusz plotting app and library - http://home.gna.org/veusz http://home.gna.org/veusz - it has a nice interactive GUI, is python scriptable and extensible via plugins. It's based around Python, Qt and numpy. It can export to PDF, EPS, SVG and bitmap formats.

Ciantic 11y ago

This is just a throw, I've not tried. But Bokeh creates Canvas. There is a Canvas 2 SVG https://github.com/gliffy/canvas2svg https://github.com/gliffy/canvas2svg JavaScript library already made.

bigreddot 11y ago

I definitely empathize. It's a technically challenging problem, there are lots of competing important priorities, and we are a small team that can only humanly accomplish so much work at once. Our vision for Bokeh was to be able to create and share interactive data applications with a minimum of python and little-to-no JS and "web tech" required. We are closing in on that vision, so it is my fervent hope (and intention) that we'll be able to turn to this important priority soon as well.