“no such column” after adding a field to the model

Question

environment DJANGO VERSION 1.9 Python 2.7.6

I added a field (scores) to a model class in models.py like this

from django.db import models
from django         


        

Answer 1

Those migration issues bother me too when i have to update my app with new columns, so i've made a small bash with some python to search for troublesome columns, remove them, migrate, re-add them and migrate again.

Here is the resolve.sh that calls python and migrations (has to be placed in the same folder as your manage.py file, like the .py files) :

python manage.py shell < check.py

read -p "Does it seem ok ? (y for ok/n for remigrate)" migrateok

if [ $migrateok = "n" ]
then
    python manage.py shell < rm_badcolumns.py

    python manage.py makemigrations
    sleep 1
    python manage.py migrate

    python manage.py shell < resume_badcolumns.py

    python manage.py makemigrations
    sleep 1
    python manage.py migrate

    echo It should work now..
else
    echo No changes were made..
fi

check.py checks if there are issues with some models in the database :

from .models import *
import os, sys
from shutil import move
from django.db import connection

tables = connection.introspection.table_names()
seen_models = connection.introspection.installed_models(tables)

errorColumn = []
is_fine = "EVERYTHING LOOKS FINE!"
#looping through models to detect issues
for model in seen_models:
    try:
        print(model.objects.all())
    except:
        is_fine = "WARNING SOME MODELS ARE CORRUPTED"

print(is_fine)

This check will tell the user if some models are not sync with your database.

If the user chooses to "re-migrate", it calls rm_badcolumns.py :

from .models import *
import os, sys
from shutil import move
from django.db import connection


tables = connection.introspection.table_names()
seen_models = connection.introspection.installed_models(tables)

errorColumn = []

for model in seen_models:
    try:
        model.objects.all()
    except:
        errorColumn.append(str(sys.exc_info()[1])[30::]+' =')
        #this weird expression just get the column that causes trouble
        #you may have to adapt indexes depending on error output in exc_info

os.chdir('.//')

#removing columns from models.py based on the error pattern
def replace(pattern, subst):
    with open('models_sub.py','w') as fout:
        with open('models.py','r') as models:
            for line in models:
                fout.write(line.replace(pattern, subst))

    os.remove('models.py')
    move('models_sub.py','models.py')

#applying this func to each error, and commenting out columns with an error mark to catch them back later
for errorStr in errorColumn:
    replace(errorStr, '#error_mark '+errorStr)

print 'Removing troublesome column for re-entering..'

Then resolve.sh will start another migration. We uncomment the troublesome columns with resume_badcolumns.py to get them migrated again :

from .models import *
import os, sys
from shutil import move
from django.db import connection


os.chdir('.//')

#same search and replace func but returning True if sthg was replaced, and False otherwise
def replace(pattern, subst):
    a = False
    with open('models_sub.py','w') as fout:
        with open('models.py','r') as models:
            for line in models:
                fout.write(line.replace(pattern, subst))
                if pattern in line:
                    a = True
    os.remove('models.py')
    move('models_sub.py','models.py')

    return a

a = True
#while error_marks are replaced, we go on looping
while a:
    a = replace('#error_mark ', '')

print 'Re-adding troublesome columns..'

We make a final migration and everything should be fine.

This is a kind of heavy and dirty artillery for a small problem but i wanted it to be automatized. A few points may need some rework, such as the way i identify which column is not synced, and i'd appreciate comments about this.