On Tuesday, August 14, 2012 2:05:25 AM UTC-4, Jeroen Demeyer wrote:
>
> What measure of "time" is this using? Unless it's wall time, your table
> below doesn't prove anything.
>
The python profiler measures wall time. Indeed this is the quantity that
you usually want to minimize, so its just c
On 2012-08-14 02:07, Volker Braun wrote:
> In any case, here is another data point. Interestingly, os.path.join is
> about as fast as posix.lstat. I'd take that as another piece of evidence
> that filesystem caches are incredibly well optimized.
What measure of "time" is this using? Unless it's wa
Doctests that ensure that certain modules are not imported upon startup are
also fine with me.
In any case, here is another data point. Interestingly, os.path.join is
about as fast as posix.lstat. I'd take that as another piece of evidence
that filesystem caches are incredibly well optimized.
On Mon, Aug 13, 2012 at 9:24 AM, Volker Braun wrote:
> On Monday, August 13, 2012 11:36:33 AM UTC-4, Robert Bradshaw wrote:
>>
>> Shouldn't everything needed by the sage notebook (already) be in
>> sagenb.* anyways?
>
>
> Yes, but thats not what I mean. For example, right now we import
> sage.grap
On Monday, August 13, 2012 11:36:33 AM UTC-4, Robert Bradshaw wrote:
>
> Shouldn't everything needed by the sage notebook (already) be in
> sagenb.* anyways?
Yes, but thats not what I mean. For example, right now we import
sage.graphs.graph_editor upon startup. And the graph editor imports the
Shouldn't everything needed by the sage notebook (already) be in
sagenb.* anyways? This should be fairly explicit, no need to move
things around further. +1 to only importing notebook() (which lazily
loads what is needed on being called).
On Mon, Aug 13, 2012 at 6:32 AM, Volker Braun wrote:
> I w
I would propose that we go one step further:
* Install all notebook dependencies into a separate directory
* only import notebook() from the sagenb on startup
* The notebook() function adds notebook dependencies to sys.path before
importing its prerequisites
This would force the notebook t
On Sun, Aug 12, 2012 at 12:51 PM, Volker Braun wrote:
> For comparison, here is sage-5.1.rc1. It looks like the notebook update
> pulled in two slow modules: flaskext.babel and pytz. And we now have a
> longer sys.path (from 26 to 39 entries), which makes module loading overall
> slower. Especiall
I get around 2 seconds on my MacBook Pro (non-SSD) that's a couple years old.
FWIW, sage -c is consistently 10% faster than `echo "" | sage`.
$ time sage -c 'print 3+2'
5
real0m1.809s
user0m1.158s
sys 0m0.705s
$ time echo "" | sage
---
