[IPython-User] [notebook] centered equations

Fernando Perez fperez.net@gmail....
Sat Apr 27 00:50:53 CDT 2013

On Fri, Apr 26, 2013 at 9:55 PM, Brian Granger <ellisonbg@gmail.com> wrote:
> I agree that there is soms serious history and weight towards centered
> latex equations.  But there are a few features that makes the notebook
> quite different from traditional latex in its visual language:
> 1) Traditional latex document have a relatively fixed aspect ratio
> that is much taller than wide (Legal, Letter, etc.).  Because of the
> dimensions of modern monitors, the aspect ratio of the typical
> notebook is opposite: much wider than tall - but also not fixed.
> 2) While traditional latex uses left or fill justify for paragraphs,
> the narrow/tall aspect ratio means that visually most paragraphs
> appear to be centered on the page.  In combination with the centered
> equations and figures, this means that all content in a math heavy
> book/journal is essentially centered on the page.  In contrast, most
> content in the notebook is left justified and strongly so: code (even
> 80 column code only takes up a small amount of the full notebook
> width), output, images, even Markdown (unless paragraphs are very long
> or the aspect ratio is narrowed).
> 3) The visual flow of a traditional latex document is down the middle
> of the page.  Because of the way code is entered interactively and
> inputy/output is displayed with prompts, the visual flow of the
> notebook is down the left wide of the notebook area.
> In my mind, these differences cause centered equations to look really
> out of place in the notebook unless:
> * The aspect ratio of the notebook area is narrowed artificially
> * There are long blocks of markdown text that uniformly fill
> horizontal space - creating the illusion of visually centered text
> like latex.
> * The code and output present in the notebook has significant
> horizontal weight that shifts the visual flow more towards the center
> of the page.
> IOW, if the notebook page+text+content take on the visual dimensions
> that regular latex documents have, equation centering makes *perfect*
> sense.  Otherwise, I don't think they do make sense.  At most, we
> should make this easier to configure on a per notebook level -
> preferrably with embedded CSS so it follows the notebook to nbviewer,
> etc.  But I do think the default is correct on this one.

Mmh, still not convinced in the slightest.  There are a couple of
assumptions I'm not sure about:

- the idea that the notebook window is automatically very wide. That's
not mandated anywhere: users can have narrow windows for improved
readability, and I suspect many people who work extensively with the
notebook probably do so, because reading unbroken lines that are many
hundreds of characters long is very painful.  I know I don't use a
full-widescreen notebook browser ever (unless giving a talk on a tiny
1024x768 projector with big fonts), since the readability is awful.

- (3) above is true only for the plain (and awful) latex 'article'
class.  Many other classes use visual layouts that are less 'big down
the middle and lots of whitespace on the sides'.  Pretty much every
journal and book uses a layout that makes more efficient use of page
whitespace than the default latex 'article' (though they are certainly
by and large block-centered.

I still think that the current default is the wrong choice.  We
certainly need to improve configurability, but we should also strive
for optimal defaults.  This one, I happen to think is not optimal,
except in one case: when the user is working in a mode so hideously
ugly (super-wide screen) that they've already jettisoned aesthetics
overboard in the first place, so there's nothing really to worry about
anymore.  For any user working with a sensible browser width, the
current default is *precisely* the wrong choice, so we're rewarding
bad setups and punishing sensible ones.

> [as a separate note, I think there are other visual problems with the
> overly wide aspect ratio of the notebook and that we should think
> about that]

Yup, that's certainly true.  That post from a while ago on typography
pointed out a number of valid issues.



More information about the IPython-User mailing list