Re: [sqlite] Format change to fts2 module.

2006-11-29 Thread Ralf Junker
Hello Scott Hess,

>http://www.sqlite.org/cvstrac/tktview?tn=2046 should fix this for fts1 and 
>fts2.

I have just tested them in both fts1 and fts2 and the reported problems no 
longer show! Many thanks for the fixes!

Please allow me to report some (compiler-independent) compiler warnings about 
fts:


* fts1.c:

Variable i is never used in function static char *firstToken(char *zIn, char 
**pzTail).

Variable j is never used in function static int parseSpec(TableSpec *pSpec, int 
argc, const char *const*argv, char**pzErr)


* fts2.c:

Both warnings above also apply to fts2.


* SQLITE_EXTENSION_INIT1

If I compile both fts1 and fts2 into the same executable with

  -DSQLITE_CORE=1
  -DSQLITE_ENABLE_FTS1=1
  -DSQLITE_ENABLE_FTS2=1

I receive a linker warning that sqlite3_api is defined in both fts1.c and 
fts2.c.

The warning goes away if I remove SQLITE_EXTENSION_INIT1 from both units. I 
might be wrong, but maybe this is not necessary if the library is compiled with 
SQLITE_CORE=1?

Regards,

Ralf 


-
To unsubscribe, send email to [EMAIL PROTECTED]
-



Re: [sqlite] Format change to fts2 module.

2006-11-28 Thread Scott Hess

http://www.sqlite.org/cvstrac/tktview?tn=2046 should fix this for fts1 and fts2.

-scott

On 11/21/06, Ralf Junker <[EMAIL PROTECTED]> wrote:

Hello Scott Hess,

>>Not directly related to the format change, but did you have a chance
>>to look at ticket #2046?
>
>Yes, I took a look at it - I _believe_ that it's not actually an issue
>with fts2, because the sequence of virtual-table calls don't really
>make sense.  I'm going to try implementing a simpler virtual table to
>isolate the problem better.

This is great news - especially since this problem is annoying me a lot. Please 
let me know if I can be of any help.

Ralf


-
To unsubscribe, send email to [EMAIL PROTECTED]
-




-
To unsubscribe, send email to [EMAIL PROTECTED]
-



[sqlite] Format change to fts2 module.

2006-11-13 Thread Scott Hess

http://www.sqlite.org/cvstrac/chngview?cn=3511

This changes the storage to delta-encode docids, rather than storing
them in absolute form.  Unfortunately, since this changes the format
of the backing data, if you've been experimenting with fts2 databases,
you'll have to regenerate them (see the email snippet below).

I'm going to try to get a couple other format-changes checked in this
week, in the hopes of getting the fts2 file format frozen.

Thanks,
scott


On 10/12/06, Scott Hess <[EMAIL PROTECTED]> wrote:

Caveats: It uses a new storage format, and there is currently no
support for "upgrading" a table from fts1 to fts2.  Additionally, I
have a number of follow-on changes which will be going in over the
next couple weeks, some of which are likely to break data
compatibility.  I'm just tossing this out here in case people are
interested in experimenting with it.


-
To unsubscribe, send email to [EMAIL PROTECTED]
-