* Vaclav Petras <wenzeslaus@gmail.com> [2018-08-23 19:09:47 -0400]:
On Thu, Aug 23, 2018 at 5:40 AM Nikos Alexandris <nik@nikosalexandris.net>
wrote:
Is there an elegant alternative?
As for your script design/coding approach:
You were concerned about nice history at one point. You can wrap it, not
just in a plain script, but in a module.
There are three things in mind, in order:
Provenance (as you name it below),
then re-usability/reproducibility
lastly legibility.
Then you can hide any particulars
of the implementation and write proper history in the module. This
avoids changing history which would lead to history not being
representative of what actually happened (provenance is the keyword
here).
Changing, or hiding, history is not wanted. As elsewhere, also in this
case, intermediate map names of some processing, get a prefix
using `script.core.tempfile()` [0], to get them easily removed at exit.
Removed if the user does not request for their output, which is an
option. If the user requests for their output, they are renamed to the
requested output name, so as to no be picked up by the clean up
`atexit.register(cleanup)` call that does this job.
[0] https://grass.osgeo.org/grass75/manuals/libpython/script.html?highlight=tempfile#script.core.tempfile
Specifically,
if((tmp.22599.18.tmp.22599.11.tmp.22599.6.recreation_potential == 1
&& tmp.22599.19.opportunity == 1), 1,
if((tmp.22599.18.tmp.22599.11.tmp.22599.6.recreation_potential == 1
&& tmp.22599.19.opportunity == 2), 2,
if((tmp.22599.18.tmp.22599.11.tmp.22599.6.recreation_potential == 1
&& tmp.22599.19.opportunity == 3), 3,
if((tmp.22599.18.tmp.22599.11.tmp.22599.6.recreation_potential == 2
&& tmp.22599.19.opportunity == 1), 4,
if((tmp.22599.18.tmp.22599.11.tmp.22599.6.recreation_potential == 2
&& tmp.22599.19.opportunity == 2), 5,
if((tmp.22599.18.tmp.22599.11.tmp.22599.6.recreation_potential == 2
&& tmp.22599.19.opportunity == 3), 6,
if((tmp.22599.18.tmp.22599.11.tmp.22599.6.recreation_potential == 3
&& tmp.22599.19.opportunity == 1), 7,
if((tmp.22599.18.tmp.22599.11.tmp.22599.6.recreation_potential == 3
&& tmp.22599.19.opportunity == 2), 8,
if((tmp.22599.18.tmp.22599.11.tmp.22599.6.recreation_potential == 3
&& tmp.22599.19.opportunity == 3), 9)))))))))
will be best represented as:
if((potential == 1 && opportunity == 1), 1,
if((potential == 1 && opportunity == 2), 2,
if((potential == 1 && opportunity == 3), 3,
if((potential == 2 && opportunity == 1), 4,
if((potential == 2 && opportunity == 2), 5,
if((potential == 2 && opportunity == 3), 6,
if((potential == 3 && opportunity == 1), 7,
if((potential == 3 && opportunity == 2), 8,
if((potential == 3 && opportunity == 3), 9)))))))))
Making it more meaningful and useful is the point.
By the way, I am still not sure how the combinations done by `r.cross`
will look like. The order of the end categories [1, 9] matters. To test,
perhaps easy to solve with an `r.categories` call and a proper set of
rules.
Some 'tmp' parts, in the names above, are prototyping leftovers,
and may be simply unnecessary. I have to correct/fix.
The question, anyhow still valid, is: what is the cost for "provenance"
if these prefixes are squashed? If the answer is "none" or "almost
none", then why not allow this squashing?
Are the design choices (program) made so far the right ones? Perhaps
they can be improved/corrected, and questions like the above can be
avoided altogether. I will look into this.
Further: what does it mean for "reproducibility"? Will it actually help
to understand better what is what, trace errors or support improving an
analysis? Perhaps, improve the algorithm/module itself?
As for legibility, I think it is as important as any other part of
programming. Spend less time in reading, more in thinking what can be
done with this tool.
If you actually publish the module or not that's a different question.
I, and my client, will publish it. It's mentioned here
https://grasswiki.osgeo.org/wiki/GRASS_GIS_Community_Sprint_Autumn_2017#In_person
and then here
https://grasswiki.osgeo.org/wiki/GRASS_GIS_Community_Sprint_Bonn_2018#In_person.
It's not yet there. There have been modifications in the underlying
algorithm(s). And there are processes to deal with licensing and
distribution.
We are doing our best.
Your script/code is needed to reproduce the data in anyway since you
want to change history and additionally it would be reasonable to use
your script/code to actually reproduce, so then the only difference is
number of files and interface of the scripts (some custom solution
versus module).
Splitting your code would also make sense because you are saying it is
quite long.
It does/is planned. It's been a long prototyping of the implementation.
Vaclav
Thank you Vaclav,
as usual your messages help a lot, in many directions.
Grateful to be (feel) part of this community,
Nikos
