Tuesday, October 13, 2015

First test results for #505

Working on #505 (Import SEPA statements from the bank).

I tried to import a real SEPA data file in a customer’s test site. A file with –theoretically– the same structure as the example file.

Lino reports the following error message (in a an alert box, i.e. probably some exception was raised somewhere):

Data truncated for column ‘ref’ at row 1

I added a temporary debugger line in lino_cosi.lib.sepa.camt:

for node in root[0][1:]:
    dd.logger.info("20151013 parsing %s for statement", node)
    statement = self.parse_statement(ns, node)

which caused a few dozen of lines like this:

201510-13 07:32:56 INFO camt : 20151013 parsing <Element {urn:iso:std:iso:20022:tech:xsd:camt.053.001.02}Stmt at 0x7f9ded9e7908> for statement

to appear in the log. So AFAICS it happens somewhere during CamtParser.parse_statement(),

But we don’t know for which statement. I thought that the xml file does not contain any newlines, that’s why the error message says “at row 1”. So I used “xmllint –format” to prettify the XML. But the message remains the same. Hm…

About the civil status

I did more than one hour of research, analysis and documentation for #536 (État civil : réduire les choix). See CivilState.

Commit and push.

Then I increased version of Lino Welfare to 1.1.26 and wrote two migrators (because Chatelet and Eupen have different database fields on Client). Hm. This was not elegant. This situation was an example where data migration à la Lino is not optimal. But after some thinking I noticed that I can maybe avoid writing these migrators by overriding the to_python method! Even though I wrote it myself, Lino happens to surprise me now and then! Of course this is yet to be confirmed by reality. Before doing an upgrade on a production site I’ll need to verify whether it works.

Oops, I also noted only now that the translation of “État civil” is not “civil state” but either “civil status” or –probably even better– “marital status”.

I read about the International Commission on Civil Status

and saw their document Guide pratique international de l’état civil BELGIQUE

First meeting with Märten

Today I met Märten and we had a brainstorming about the question “How to find Lino users in Estonia?” And we agreed that Lino has one important show stopper in Estonia: #587.

Python shell without arrow keys and history

A disturbing issue fixed en passant: When the arrow keys don’t work in a Python shell, then it might be simply because readline is not installed. To fix it:

$ pip install readline

But you will also need one or several of the following:

$ sudo apt-get build-dep libreadline-dev
$ sudo apt-get install libreadline-dev
$ sudo apt-get install libncurses5-dev

Migrating civil statuses

I tried whether abovementioned trick of overriding the to_python method works.

Here is a script that I used to report the distribution of clients:

from lino.api.shell import *
from django.utils import translation
CS = rt.models.pcsw.CivilState
ses = rt.login()
num = 0
with translation.override('de'):
  for cs in CS.objects():
  #for k in '13 18 21 22 33'.split():
  #  cs = CS.get_by_value(k)
    qs = rt.models.pcsw.Client.objects.filter(civil_state=cs)
    print cs.value, cs, ":", qs.count(), "Klienten"
    num += qs.count()
  qs = rt.models.pcsw.Client.objects.filter(civil_state='')
  print "Ohne Angabe:", qs.count(), "Klienten"

Output of above script in the current production database:

10 single :  301 Klienten
13 Ledig zusammenwohnend :  41 Klienten
18 Ledig mit Kind :  35 Klienten
20 married :  96 Klienten
21 Verheirated (allein lebend) :  8 Klienten
22 Verheirated (mit anderem Partner lebend) :  18 Klienten
30 widowed :  11 Klienten
33 Verwitwet (zusammenwohnend) :  1 Klienten
40 divorced :  68 Klienten
50 separated :  23 Klienten
Ohne Angabe: 5832 Klienten

Output in test database after migration:

10 single :  336 Klienten
20 married :  96 Klienten
30 widowed :  12 Klienten
40 divorced :  68 Klienten
50 separated :  23 Klienten
51 separated_de_facto :  26 Klienten
60 cohabitating :  41 Klienten
Ohne Angabe: 5832 Klienten

This confirms that the lino_welfare.modlib.pcsw.choicelists.CivilState.old2new() method has been applied correctly during migration. (e.g. the 41 clients who had status 13 have been converted to cohabitating). Lino is cool!

More test results for #505

I started a Django shell in a test environment on a customer server and ran the following commends there:

from lino.api.shell import *
ses = rt.login()

This produced the same error message. Which is actually good. I remember that I once had problems with lxml (and even stopped to use it) because it simply did not work under multi-threaded mod_wsgi. But it seems as if that’s not the problem here.

Instead of continuing to try to read the real data, I added a test case lino_welfare.projects.eupen.tests.test_import_sepa and a fictive demo data file.

The new test case fails (as predicted) because the logic for processing the incoming data is currently just a proof of concept. Here is the traceback which i get:

File "lino_cosi/lib/sepa/models.py", line 74, in run_from_ui
  self.import_file(ar, filename)
File "lino_cosi/lib/sepa/models.py", line 100, in import_file
  account = Account.objects.get(id=1)
File "site-packages/django/db/models/manager.py", line 151, in get
  return self.get_queryset().get(*args, **kwargs)
File "site-packages/django/db/models/query.py", line 310, in get
DoesNotExist: Account matching query does not exist.

I started to make changes in lino_xl.lib.sepa.models to get the test pass. The rule here is that there should be one Account for every incoming IBAN. If no Account exists, import_file() should create an orphaned account (with empty “partner” field since you don’t know the partner). The Account.partner field is now to null=True and blank=True. (I currently imagine that the end-users will have to assign partners manually since real name and the bank’s name can differ.)

Here I got stuck and went to bed. Hamza, please look at my changes. You might have more experience than I with this CAML parser. I am surprised to discover that the local_account field of a BankStatement can be empty…