Re: pysqlite design decisions

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

glyph-TyWPi3/[email protected] wrote:
|  What I really want to return is a mutable octet array, but Python's
| array module is not widely used or understood (and is missing many
| useful convenience functions that are present on str).

We'll have to agree to disagree.  I still think str is really bad object
to store blobs in.  It is ok to temporarily convert behind the scenes
into str because of what various APIs require etc.

| That's an interesting motivation.  In retrospect, I personally would
| still have made it explicit, rather than based on the string's content,
| what type apsw would return for strings (especially since if you had a
| database manipulated with the Unicode version and you attempted to open
| it with the ANSI version... what happens?) but I suppose hindsight is
20/20.

It all works perfectly because ASCII is a subset of UTF-8.  So providing
you never put any non-ASCII data in, you would never get non-ASCII out.

Sure you run a Unicode version of BitPim and then enter Unicode
characters, save them to the database, and then reopen with the ANSI
version of BitPim.  At that point apsw would return Unicode strings for
those particular fields and then ANSI wxPython would say it has no idea
what to do with the Unicode strings.  But if you could run the Unicode
version anyway, why bother with the ANSI version.

This scheme only really has trouble if there are other encodings that
are a subset of UTF-8.  One example you gave is punycode.  But it wasn't
in use much by Python people back then, nor was that much code Unicode
aware.  Hence the decision to make it work simply for all-ASCII code as
well as Unicode aware code.

And now that times have changed, I am fixing the next release to only
return Unicode :-)

Roger
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.5 (GNU/Linux)

iD8DBQFFeR80mOOfHg372QQRAtzKAKCUAbbuyFi8IXYJq0tEGTaD3kav2ACfR+GS
YRrJMZE69Tj6Qd+v6AzHBUY=
=idV2
-----END PGP SIGNATURE-----