Plugin row_factory values in sqlite3.connection

[DaltonST]

After searching this plugin forum, the sqlite3 documentation at https://docs.python.org/2/library/sq...on.row_factory, and calibre>src> , I am left with the single option of posting here to ask a simple question that is amazingly not answered in detail in the above documentation: What are the valid values for row_factory that will not return an error, especially the common "Float type is required" error.

My code at the moment is:

row_factory = sqlite3.Row
my_db = sqlite3.connect(path,row_factory,detect_types=sqli te3.PARSE_DECLTYPES|sqlite3.PARSE_COLNAMES)


There are a few values that failed, although sqlite.py shows them to be conceptually valid:

1. row_factory=(lambda cursor, row : list(row))
2. row_factory=None
3. row_factory=Connection

In standalone Python 2.7 using Sqlite3 on Windows 7, a simple "db = sqlite3.connect(dbpath)" works just fine. I am in need of the secret handshake that Calibre wants me to provide. Can you help?

Thank you in advance for your Calibre wisdom.

[davidfor]

All of mine use

with closing(sqlite.connect(self.device_database_path())) as connection:
    connection.text_factory = lambda x: unicode(x, "utf-8", "ignore")
    connection.row_factory = sqlite.Row

"self.device_database_path()" returns the full path to the database.

I don't use the row_factory everywhere. The code I was working with initially didn't and just used numeric indices into the results. I'm now using a mix depending on when I wrote the code and how many columns are in the result. If I have to take my shoes of to count, I use the names

[kovidgoyal]

Note that calibre does not use the python sqlite3 module, it uses apsw. And I never use row factories, they are an unnecessary performance drain.