Christophe Simonis @ Tiny

OpenERP 5.0.0-alpha is out...

2008-11-04T00:00:00.006+01:00

Hi, folks !

The alpha version of the OpenERP 5.0.0 is now available for download !!

If you found bugs, please fill a bug report on launchpad.

Even this version is full of cool new features, there are a few to finish for the final version:

Better calendar view in the GTK client
Process review and improvement
Editable gantt charts
Bugfix in localisation of reports
Integration with the official documentations (books)
Auto-installer for OpenERP-Web-Client
Some small improvements that will be described later (keep an eye on the OpenERP planet)

Please post your comment on the OpenERP forum: http://openerp.com/forum/post26162.html#26162

перево́д

2008-09-25T10:18:00.005+02:00

No, this blog hasn't been hacked!
The title means "translation" in Russian.

You guess it, I'll talk about the internationalization (abbreviated to i18n) of OpenERP.

Contrary to the 4.2.x version, the translations are now done by module. So, instead of an unique i18n folder for the whole application, each module has its own i18n folder. In addition, OpenERP can now deal with .po files as import/export format. The translation files of the installed languages are automatically loaded when installing or updating a module. OpenERP can also generate a .tgz archive containing well organised .po files for each selected module.

The reason why we did these changes is that we now use our launchpad project for manage the translations. Even if there are still some files awaiting to be reviewed by the launchpad team, you can start translate.

The ball is in your court now: translate !

New new domain notation

2008-08-14T14:38:00.005+02:00

Last week, I told you about the new domain notation. After a little peer reviewing, it appears that this notation is not as good as it seems. The inability to negate an expression (and not just a leaf) is more problematic that I have imagined. Thus, after a tiny discussion with Fabien, we decided to implement a (nearly) true polish notation for the domains. From now, the domains look more like the older ones that in the last week notation. This means there is no more extra parenthesis.
In addition of the classic leafs, you can add operators.

« Wait... what do you name leafs? »
A leaf, is a tuple of 3 elements composed of:

a field name
an operator (as string)
a value

Example:


('active', '=', True)
('foo', '!=', 'bar')
('id', 'in', [7, 13, 42, 666])
('id', '=', False)
('name', 'like', 'tiny')

« And which operators are allowed? »
Like the last week notation, the AND ('&') and OR ('|') operators are allowed . Unlike the last week notation, the arity of these operators is fixed to 2.
And yes, the one you have all been waiting for is now in: The NOT ('!') operator. This one has an arity of 1.
For compatibility with existing domains, the AND operator is applied by default.

« So, what looks like this new new notation? »
Some examples:


['&', ('active', '=', True), ('value', '!=', 'foo')]
['|', ('active', '=', True), ('state', 'in', ['open', 'draft'])
['&', ('active', '=', True), '|', '!', ('state', '=', 'closed'), ('state', '=', 'draft')]
['|', '|', ('state', '=', 'open'), ('state', '=', 'closed'), ('state', '=', 'draft')]
['!', '&', '!', ('id', 'in', [42, 666]), ('active', '=', False)]

PS: Tanks to Najlâa and Olivier for the review.

Python: reverse enumerate

2008-08-14T12:12:00.007+02:00

This morning, I was needed to iterate trough a list from the end to the start and getting the index in addition to the value.

For getting the index, it's easy: we use the builtin function enumerate.

>>> a = ['a', 'b', 'c']
>>> it = enumerate(a)
>>> it
<enumerate object at ...>
>>> it.next()
(0, 'a')
>>> it.next()
(1, 'b')

For iterate in the other sens, it's also easy; there is the builtin function reversed.

>>> a = ['a', 'b', 'c']
>>> it = reversed(a)
>>> it
<listreverseiterator object at ...>
>>> it.next()
'c'
>>> it.next()
'b'

Therefore, getting the index with the element in the reverse order would be easy as the combination of the two functions:

>>> a = ['a', 'b', 'c']
>>> it = reversed(enumerate(a))
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
TypeError: argument to reversed() must be a sequence

Argh! We must transform the enumerate object to a list before passing it to reversed:

>>> a = ['a', 'b', 'c']
>>> it = reversed(list(enumerate(a)))
>>> it.next()
(2, 'c')

It works.
Yes, it works, but there is a problem. It's not evident for everybody but this technique isn't optimized. First we create an iterator object with enumerate, then we convert this object to a list before recreate another iterator object. With a tiny list of three elements like in my examples it's not a pain, but with big lists, it consume a lot of memory by doing a copy of the initial list with the indexes.

The solution is no more complicated but don't use the method enumerated. We simulate his behavior with the method izip from the itertools module and the method xrange. Both methods create iterator, which is our goal.
The itertools.izip method take n iterators in parameters and yield the m^th element of each iterator. Example

>>> from itertools import izip
>>> a = ['a', 'b', 'c']
>>> b = ['d', 'e', 'f']
>>> it = izip(a, b)
>>> it.next()
('a', 'd')
>>> it.next()
('b', 'e')

Thus, we can rewrite the enumerate method like this:

>>> a = ['a', 'b', 'c']
>>> it = izip(xrange(len(a)), a)
>>> it.next()
(0, 'a')
>>> it.next()
(1, 'b')

Now, to reverse the order we just reverse all the izip arguments

>>> a = ['a', 'b', 'c']
>>> it = izip(xrange(len(a)-1, -1, -1), reversed(a))
>>> it.next()
(2, 'c')
>>> it.next()
(1, 'b')

To finish we hide this behind a lambda to have a more readable code...

>>> reverse_enumerate = lambda l: izip(xrange(len(l)-1, -1, -1), reversed(l))
>>> a = ['a', 'b', 'c']
>>> it = reverse_enumerate(a)
>>> it.next()
(2, c)

This simple code has been implemented into openobject-server with the revision 964.

New domain notation

2008-08-05T11:13:00.005+02:00

During the last Community Days, somebody (it was Joël from C2C, I think) complain about the simplicity of the domains. For example, you can't include a OR condition, or negate the domain.

Negate a domain is a false problem: you may just invert the operator. So, [('active', '=', 1)] become [('active', '<>', 1)]

But, for the OR operator, the problem is really there. An implicit AND operation is made between each part of the domain. [('active', '=', 1), ('parent_id', '=', 42)] is translate into a SQL query that look like WHERE active = 1 AND parent_id = 42

Yes, adding a OR condition, is not a bad idea, but, how to include it without breaking all the domains already written ? After some discussions, we decide to uses a pseudo polish notation, always with a implicit AND operator. Now you can prefix domain parts with a '&' or a '|' and group part using parenthesis.

Here are some example of domains and the resulting SQL where clause:

[('foo', '=', 'bar')]
foo = 'bar'

[('id', 'in', [1,2,3])]
id in (1, 2, 3)

[('field', '=', 'value'), ('field', '<>', 42)]
( field = 'value' AND field <> 42 )

[('&', ('field', '<', 'value'), ('field', '>', 'value'))]
( field < 'value' AND field > 'value' )

[('|', ('field', '=', 'value'), ('field', '=', 'value'))]
( field = 'value' OR field = 'value' )

[('&', ('field1', '=', 'value'), ('field2', '=', 'value'), ('|', ('field3', '<>', 'value'), ('field4', '=', 'value')))]
( field1 = 'value' AND field2 = 'value' AND ( field3 <> 'value' OR field4 = 'value' ) )

[('&', ('|', ('a', '=', 1), ('b', '=', 2)), ('|', ('c', '=', 3), ('d', '=', 4)))]
( ( a = 1 OR b = 2 ) AND ( c = 3 OR d = 4 ) )

[('|', (('a', '=', 1), ('b', '=', 2)), (('c', '=', 3), ('d', '=', 4)))]
( ( a = 1 AND b = 2 ) OR ( c = 3 AND d = 4 ) )

Don't hesitate to ask questions if you don't understand.

Stay tuned.

PS: Thank you to Stéphane for the base code of the tree.

Now, you know when there are attachments !

2008-06-20T10:04:00.005+02:00

Since a few days, with the trunk version, you can easily know if there are attachments on a resource. The attachment button is different and the number of attachments is shown.

And it's also work in eTiny !

Stay tuned !

Welcome Tiny

2008-06-08T21:55:00.001+02:00

Since the 14th may, I'm 26 years old.
But this is also when I began working for Tiny.
And this is exactly what I'm going to talk about on this blog: how I improve OpenERP.
I'll work on support (bug fixing) but also on code improvement.