[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Blends and (astronomy) meta-packages



Andreas Tille <andreas@an3as.eu> writes:
>> The nice thing about python libs is that they usually serve as end-user
>> applications (though ipython), and also as development packages. The
>> ease of conversion between an interactive analysis to a programs is IMO
>> what makes Python so successfull (and therefore maybe an
>> python-astronomy package would be useful).
>
> There exist packages that contain applications written in Python where
> the user does not really need to know in what language it is written
> in (as for any user application). As far as I understand packages
> featuring such applications should not feature the 'python-' prefix in
> front of the package name.

I don't mean these.

> Python *modules* are useful to develop Python applications and usually
> these come in a python-* and a python3-package.  May be there is
> something in between but that split can be drawn through close to all
> Python packages I know.  And than there might be exceptions which
> serve a purpose inbetween and you can create from one source package
> pure applications to run at the users computer and modules that help
> to program similar applications.  I would split up these parts of one
> source package into different binary packages and put them into
> different Blends tasks - one in the user oriented task and one in a
> development task.

It is quite common on astronomy (and other science fields, like high
energy physics) to start a specific "shell" and then do you analysis
there. In astronomy, you have a number of these "shells": IRAF,
ESO-MIDAS as (wanna-) free versions, or IDL as a commercial one.

Recently, the most common shell is "ipython", which even comes with a
nice graphical "notebook" [1]. So, one just starts ipython does an
"import astropy" and interactively runs the analysis [2].

Therefore, the astronomy python modules are not strictly "development",
but they are in fact extensions to our analysis shells, and it makes
sense to have them separated (answering the question "which packages can
I use in my IPython notebook?")

>> If it would be created independently, would
>> it be easy to integrate as a blends task if I change my mind?
>
> There are tools (package blends-dev) that turn a simple tasks file into
> a metapackage.  If you have a manually craftet metapackage you can
> probably decide to copy the Depends/Recommends and Suggests into a tasks
> file if you want to turn it into a Blends task.  However, my
> recommendation to start right now with the design of *some* Astronomy
> tasks (not only for Python applications) remains and I recommend to
> study the design of other Blends tasks.

Can I use the blends-dev tasks without actually creating a blends task
but a "normal" metapackage?

Best regards

Ole

[1] http://ipython.org/notebook.html
[2] http://www.astropy.org/astropy-tutorials/UVES.html
    as example tutorial


Reply to: