Wednesday, September 13, 2017

Jane performance

I converted lino_xl.lib.clocking to disable the start_session and end_session actions via disabled_fields (for #2044, now that #2048 is done). And I set the end_time field of Session to db_index=True. En passant I moved docstrings from the code to Work time tracking (#1869).

Getting Lino Book to build on readthedocs

Wow, the Book test suite passed for the first time on both Python 2 and 3!

On readthedocs we continue to have the problem that the build gets killed: But there was a warning about

Sometimes it has the following error:

OSError: [Errno 2] No such file or directory:

And I don’t see where it comes from. There was once upon a time a module mixins in lino.modlib.system, and I discovered only recently that a .pyc file was still dangling in my working copy. And that might have caused autodoc to generate a file api/lino.modlib.system.mixins.rst

Optimizations for Lino Avanti

The insert_layout for had gone, and also the ⊕ Button to quickly add a new knowledge.

lino_xl.lib.trends.TrendEvent has no longer unique_together set to [‘subject’, ‘trend_stage’]. It can happen that a same stage occurs more than once for a given subject.

Optimizations for Lino Tera

Aufräumen in Stammdaten Klienten

Jetzt geht es ans Erfassen der Zahlungen, und da fängt es schon an zu stören, dass doch recht viele Klienten oder Zahler doppelt existieren. Das kommt scheinbar davon, dass sie in TIM nicht korrekt miteinander verknüpft waren.

Manuelle Korrektur : Enrolment.start_date ist korrekt ausgefüllt, aber diverse andere Datenfelder bei den Aktivitäten stehen konstant auf 29.07.2017 (Datum des Imports). Hier mein Skript

from import courses
for obj in courses.Enrolment.objects.all():
    obj.request_date = obj.start_date
    obj.course.start_date = obj.start_date
  • detail window of a place now shows the partners living there
  • merge action also for persons. to observe : what happens when this is used on a person who is also a client.
  • Show zip code when selecting a city (#2064).

NB: I did not yet adapt the test suites to above changes. book and welfare are failing.

User.username can not be null

During a deploy for Lino Avanti the failed because they had two “virtual” users with empty username. #2065. An empty username was perfectly allowed in Lino. It means that this user exists and can be referred to, e.g. as the author of some UserAuthored, but cannot be used for logging in. This feature was not tested however.

At first I thought it had to do with nullable charfields in general. I read this thread and removed dd.NullCharField. But still the problem didn’t go away : the dump contained two users with username None, but the database complained about duplicate username fields.

Upgraded from Django 1.11.3 to 1.11.5. Nope.

>>> from lino import startup
>>> startup("lino_book.projects.adg.settings.demo")
>>> from lino.api.doctest import *
>>> #fld = users.User._meta.get_field('username')
>>> fld = avanti.Client._meta.get_field('national_id')
>>> fld.__class__
<class 'django.db.models.fields.CharField'>
>>> fld.null
>>> fld.blank
>>> fld.unique
>>> fld.empty_strings_allowed

A new test in min1 revealed that the guilty was django.contrib.auth.base_user.AbstractBaseUser: it seems that Django doesn’t tolerate username being None. If it is None, Django sets it to “None”. What a pity! Okay, I change Lino’s username to be non blank and non nullable.

The only strange thing is that Lino did let us save both User objects with empty username.

Public money, public code!

Publicly financed software developed for the public sector should be made publicly available under a Free and Open Source Software licence. Unfortunately this is far from being reality. It’s a shame. I signed the open letter.