Path: csiph.com!v102.xanadu-bbs.net!xanadu-bbs.net!feeder.erje.net!eu.feeder.erje.net!xlned.com!feeder1.xlned.com!newsfeed.xs4all.nl!newsfeed1.news.xs4all.nl!xs4all!newsgate.cistron.nl!newsgate.news.xs4all.nl!post.news.xs4all.nl!not-for-mail
MIME-Version: 1.0
From: Matthew Lefavor <mclefavor@gmail.com>
Date: Wed, 7 Aug 2013 15:38:48 -0400
Subject: Is a Metaclass the appropriate way to solve this problem?
To: Python Mailing List <python-list@python.org>
Content-Type: multipart/alternative; boundary=001a11c3a3042b243404e360b382
Precedence: list
Newsgroups: comp.lang.python
Message-ID: <mailman.326.1375904372.1251.python-list@python.org>
Lines: 101
NNTP-Posting-Host: 2001:888:2000:d::a6
Xref: csiph.com comp.lang.python:52146

--001a11c3a3042b243404e360b382
Content-Type: text/plain; charset=ISO-8859-1

All:

Like most people, I find the whole metaclass topic pretty obscure, and I
have avoided trying to use one for a while. I am also aware of Tim Peter's
famous advice that if you have to ask whether you need a metaclass, then
you almost certainly don't. But in this case I know I am solving a problem
similar to problems that other high-profile Python modules (e.g., Django's
Model class) have solved using metaclasses.

My company is using a database whose interface is defined in a series of
JSON objects. I am writing some interface code to the database and I am
trying to hide the detail of the JSON implementation from the user.

I want the user to be able to define a class representing a database entry
for any arbitrary table, whose attributes represent database entities, like
fields or tags, with syntax something like the following:

class DataSet:
    data_set_id = DatabaseKeyField(int)
    description = DatabaseField(str)
    creation_date = DatabaseField(datetime.date)
    creation_timestamp = DatabaseField(datetime.datetime)

    def __init__(self, ds_id, description, timestamp):
        self.data_set_id = ds_id
        self.description = description
        self.creation_timestamp = timestamp
        self.creation_date = timestamp.date

I know that to create the DatabaseField objects I should be using a
descriptor. But I also want the DataSet to automatically gain methods that
will convert it into the expected JSON syntax (e.g., a __specifier__ method
that will create a JSON object with only "key" fields, and an __object__
method that will create a JSON object with all the fields and other bells
and whistles.)

My ultimate question then: How do I go about adding those kinds of methods
(which iterate through all the attributes which are Database* descriptors)?
I know that I could eschew metaclasses altogether and use a common
super-class, but this doesn't feel like the right situation for inheritance
to me. Is a metaclass going to be the cleanest and easiest-to-understand
way to solve this problem?

Thank you, all!

-MCL

--001a11c3a3042b243404e360b382
Content-Type: text/html; charset=ISO-8859-1
Content-Transfer-Encoding: quoted-printable

<div dir=3D"ltr">All:<div><br></div><div>Like most people, I find the whole=
 metaclass topic pretty obscure, and I have avoided trying to use one for a=
 while. I am also aware of Tim Peter&#39;s famous advice that if you have t=
o ask whether you need a metaclass, then you almost certainly don&#39;t. Bu=
t in this case I know I am solving a problem similar to problems that other=
 high-profile Python modules (e.g., Django&#39;s Model class) have solved u=
sing metaclasses.</div>

<div><br></div><div>My company is using a database whose interface is defin=
ed in a series of JSON objects. I am writing some interface code to the dat=
abase and I am trying to hide the detail of the JSON implementation from th=
e user.</div>

<div><br></div><div>I want the user to be able to define a class representi=
ng a database entry for any arbitrary table, whose attributes represent dat=
abase entities, like fields or tags, with syntax something like the followi=
ng:</div>

<div><br></div><div>class DataSet:</div><div>=A0 =A0 data_set_id =3D Databa=
seKeyField(int)</div><div>=A0 =A0 description =3D DatabaseField(str)</div><=
div>=A0 =A0 creation_date =3D DatabaseField(datetime.date)</div><div>=A0 =
=A0 creation_timestamp =3D DatabaseField(datetime.datetime)</div>

<div>=A0 =A0=A0</div><div>=A0 =A0 def __init__(self, ds_id, description, ti=
mestamp):</div><div>=A0 =A0 =A0 =A0 self.data_set_id =3D ds_id</div><div>=
=A0 =A0 =A0 =A0 self.description =3D description</div><div>=A0 =A0 =A0 =A0 =
self.creation_timestamp =3D timestamp</div>

<div>=A0 =A0 =A0 =A0 self.creation_date =3D timestamp.date</div><div><br></=
div><div>I know that to create the DatabaseField objects I should be using =
a descriptor. But I also want the DataSet to automatically gain methods tha=
t will convert it into the expected JSON syntax (e.g., a __specifier__ meth=
od that will create a JSON object with only &quot;key&quot; fields, and an =
__object__ method that will create a JSON object with all the fields and ot=
her bells and whistles.)</div>

<div><br></div><div>My ultimate question then: How do I go about adding tho=
se kinds of methods (which iterate through all the attributes which are Dat=
abase* descriptors)? I know that I could eschew metaclasses altogether and =
use a common super-class, but this doesn&#39;t feel like the right situatio=
n for inheritance to me. Is a metaclass going to be the cleanest and easies=
t-to-understand way to solve this problem?</div>

<div><br></div><div>Thank you, all!</div><div><br></div><div>-MCL</div></di=
v>

--001a11c3a3042b243404e360b382--