CherryPy + Database questions

[Brian Blais]

Hello,

I have more of a conceptual question about the way databases work, in a web
framework, but I will be implementing things with CherryPy and SQLAlchemy. If you
make a web form that adds rows to a database, and use sqlite as the engine, is there
a danger of a race condition if several users at the same time try to submit the
form? I want a unique row id, and make a unique identifier for the submissions. Am
I in danger of having multiple submissions given the same identifier?

I know that if I implemented things without a database, but used a flat file or
something like it, I would have that possibility, when two processes try to read and
update the file at the same time.

If you want something more specific, I can throw some code together, but I wanted to
get an answer before that because it will change the way I develop the code.


thanks,

Brian Blais

 

[Diez B. Roggisch]

Hello,

I have more of a conceptual question about the way databases work, in a
web
framework, but I will be implementing things with CherryPy and SQLAlchemy.
If you make a web form that adds rows to a database, and use sqlite as
the engine, is there a danger of a race condition if several users at the
same time try to submit the
form? I want a unique row id, and make a unique identifier for the
submissions. Am I in danger of having multiple submissions given the same
identifier?

I know that if I implemented things without a database, but used a flat
file or something like it, I would have that possibility, when two
processes try to read and update the file at the same time.

If you want something more specific, I can throw some code together, but I
wanted to get an answer before that because it will change the way I
develop the code.

Usually, that is exactly what a RDBMS gives you. See the database connector
module's threadsaftey property to see what exactly you can expect.

Diez

 

[timothy.kellogg]

Usually, that is exactly what a RDBMS gives you. See the database connector
module's threadsaftey property to see what exactly you can expect.

Diez

SQLite isn't really a DBMS at all, it's just a library. According to
their FAQ (http://www.sqlite.org/faq.html#q8) it is only threadsafe as
long as you aren't changing the same DB in separate threads (so no,
not threadsafe at all).

Hope this helps.

--Tim

 

[Diez B. Roggisch]

SQLite isn't really a DBMS at all, it's just a library.

Where is the difference? From the DB-API-side none. And there are plenty of
DBs out there that work similar, like MS Access. Show me where it is
written that a DBMS has to run as separate server process.

According to
their FAQ (http://www.sqlite.org/faq.html#q8) it is only threadsafe as
long as you aren't changing the same DB in separate threads (so no,
not threadsafe at all).

I'm not sure if I share your POV. According to the DB API threadsafty
attribtue, it's safe to have multiple threads working on a single DB if
each has his own connection. Which is a common idiom anyway.

And AFAIK e.g. TurboGears doesn't perform any special measures to
synchronize cherrypy worker threads when using sqlite, albeit I'm not 100%
sure on this.

Diez