Using django ORM from web browser and from command line apps

[News123]

Hi,

I'm having a django browser application.

There's certain administrative tasks, that I'd like to perform from the
command line (cronjob or manually).
As these scripts might be huge and might consume quite some memory I'd
prefer, that they were not part of the normal application and would just
consume memory during the administrative phase.

I am not sure, whether this is a use case really being intended.

I wanted to know whether there are any precautions to take if I do this.

The one issue, that I could imagine is that Django (if I understood
correctly) performs normally database transaction based an a http request.

What would happen if I don't do anythong special in a script.
Would the entire runtime of the script be considered a transaction?

Are there any special commands that I have to use in order to indicate
when a transaction starts / stops?


Thanks in advance for suggestions.

Please look also at a related question.

 

[Ian Kelly]

Hi,

I'm having a django browser application.

There's certain administrative tasks, that I'd like to perform from the
command line (cronjob or manually).
As these scripts might be huge and might consume quite some memory I'd
prefer, that they were not part of the normal application and would just
consume memory during the administrative phase.

I am not sure, whether this is a use case really being intended.

It sounds like you probably want a custom management command:

https://docs.djangoproject.com/en/1.3/howto/custom-management-commands/

I wanted to know whether there are any precautions to take if I do this.

The one issue, that I could imagine is that Django (if I understood
correctly) performs normally database transaction based an a http request.

If you have the TransactionMiddleware enabled, yes. Otherwise the
default is to commit everything immediately.

What would happen if I don't do anythong special in a script.
Would the entire runtime of the script be considered a transaction?

The default here is also to commit everything immediately. Since
there is no HTTP request, the TransactionMiddleware does not get
invoked even if enabled. For controlled transactions you will need to
use the commit_on_success or commit_manually decorator /
context-managers:

https://docs.djangoproject.com/en/1.3/topics/db/transactions/

Cheers,
Ian

 

[News123]

Hi Ian,

It sounds like you probably want a custom management command:

https://docs.djangoproject.com/en/1.3/howto/custom-management-commands/

Didn't know this existed.
Yes this is definitely a way to go. At least for part of the commands,
that I plan to implement.
Out of curiousity: Do you know whether the imports would be executed for
each potential command as soon as I call manage.py or only
'on demand'?

If you have the TransactionMiddleware enabled, yes. Otherwise the
default is to commit everything immediately.

Good to know.

The default here is also to commit everything immediately. Since
there is no HTTP request, the TransactionMiddleware does not get
invoked even if enabled. For controlled transactions you will need to
use the commit_on_success or commit_manually decorator /
context-managers:

https://docs.djangoproject.com/en/1.3/topics/db/transactions/

Thanks this url is very clear.

I will use the commit model as required for my different tasks

 

[Ian Kelly]

Out of curiousity: Do you know whether the imports would be executed for
each potential command as soon as I call manage.py or only
'on demand'?

Off the top of my head, I don't know.

 

[News123]

Off the top of my head, I don't know.

Never mind.
WIill jsut add a print statement when I make my first test.
This should clearly show whether the module is imported all the time or
only on demand.

 

[bruno.desthuilliers]

Out of curiousity: Do you know whether the imports would be executed for
each potential command as soon as I call manage.py or only
'on demand'?

Why would you care ? Just importing the module shouldn't have any side
effect.