On Tue, 2005-07-12 at 11:01 -0700, Derek Shaw wrote:
> SQLite 3 relocates 2 addresses out of its memory bounds when it
> loads.
Can you explan in more detail what this means?
--
D. Richard Hipp <[EMAIL PROTECTED]>
as written a completely different program by the same name.
You want the "sqlite3_analyzer" that appears on on
http://www.sqlite.org/download.html
--
D. Richard Hipp <[EMAIL PROTECTED]>
ments - ever.
--
D. Richard Hipp <[EMAIL PROTECTED]>
T rowid FROM a WHERE f1=5 UNION
SELECT rowid FROM a WHERE f2=11);
Perhaps SQLite will do such rewriting automatically someday,
but probably not in the near term.
--
D. Richard Hipp <[EMAIL PROTECTED]>
t; Now execute a query using both the indexes:
>
> SELECT * FROM a WHERE f1 = 1 OR f2 = 1;
> MySQL and PostgreSQL will use the indexes here, and
> therefore return the result considerably faster.
>
Really? I would be very interested to know what
query plan MySQL and PostgreSQL u
ement
around until after the new one is ready, then use the
sqlite3_transfer_bindings() API to transfer all your bindings
from the old to the new, then finalize the old.
>
> Is it possible to get the SQLITE_SCHEMA error after the first
> sqlite3_step call, while iterating throw the rows?
>
No. SQLITE_SCHEMA will always appear immediately or not
at all.
--
D. Richard Hipp <[EMAIL PROTECTED]>
On Sun, 2005-07-10 at 12:21 -0400, [EMAIL PROTECTED]
wrote:
> IIRC, you can't create a table in an attached DB.
That's true for SQLite 2.8. But beginning with SQLite 3.0
the restriction is removed and you can create new tables in
an attached database.
--
D. Richard Hipp <[EMAIL PROTECTED]>
et it to use a wrapper class of some sort that
automates the task of rerunning sqlite3_prepare() when necessary.
--
D. Richard Hipp <[EMAIL PROTECTED]>
t cannot. That way, you can never get two threads holding
a shared lock and both trying to promote to an exclusive lock.
This avoids the deadlock scenario described above, but it also
reduces the amount of concurrency. So it is a tradeoff.
--
D. Richard Hipp <[EMAIL PROTECTED]>
nd from: http://www.piclist.com/techref/microchip/math/32bmath-ph.htm
>
> "signed 32-bit numbers have a range of -2147483648 to +2147483647
>
That's correct for version 2.8. Version 3.0 expands the INTEGER PRIMARY
KEY out to 64 bits so you have a range of -18446744073709551616 to
+18446744073709551615. Seems unlikely to overflow...
--
D. Richard Hipp <[EMAIL PROTECTED]>
nt to put $var inside
your SQL statement and let SQLite worry about escapes rather than
doing so yourself. Like this:
DB eval BEGIN
DB eval {INSERT INTO table1 VALUES($a,$b,$c,$d)}
DB eval COMMIT
--
D. Richard Hipp <[EMAIL PROTECTED]>
tion will
cause the transaction to roll back automatically.
I *hope* that is what the SQLite integration with firefox
will accomplish. But again, I don't really know.
--
D. Richard Hipp <[EMAIL PROTECTED]>
On Wed, 2005-07-06 at 09:32 -0400, [EMAIL PROTECTED]
wrote:
> "D. Richard Hipp" <[EMAIL PROTECTED]> writes:
>
> > On Wed, 2005-07-06 at 09:38 -0300, Gerson LuÃs Fontoura Vaz wrote:
> >> I'm using SQLite for the first time...
> >>
> >> We
k or to unbounded growth
in memory requirements.
--
D. Richard Hipp <[EMAIL PROTECTED]>
memory leak in version 3.2.2, we
would look into the matter right away. For 2.8.15, you can
file a bug report if you like. Somebody might investigate the
problem in a month or two.
--
D. Richard Hipp <[EMAIL PROTECTED]>
On Fri, 2005-07-01 at 20:22 -0500, Dan Wellisch wrote:
> Richard:
>
> Are you saying that the statement should be.:
>
> wxString::Format(wxT("%s%s"),zipCodeValue.c_str(), "%");
>
Yes. Or perhaps this:
wxString::Format(wxT("%s%%"), zip
), SQLITE_STATIC);
>
Do not put quotes around strings that are bound. Just insert
them directly. You are trying to LIKE against '''6008%''', not
'6008%' as you intend.
--
D. Richard Hipp <[EMAIL PROTECTED]>
On Thu, 2005-06-30 at 16:35 +0530, Ajay wrote:
> So I used query
> Insert into NEWTABLE select * from OLDTABLE order by no desc
> But it is not giving me sorted output as new table?
>
> Can you tell me where I am wrong ???
>
The ORDER BY clause on a SELECT used to insert into a table
has been
it appears to be have been addressed
> post-v3.2.1.
>
http://www.sqlite.org/cvstrac/tktview?tn=1188
I had forgotten about that one. Sorry.
--
D. Richard Hipp <[EMAIL PROTECTED]>
ou are are getting the results that
you report. I am unable to reproduce them.
--
D. Richard Hipp <[EMAIL PROTECTED]>
e, I get only a single
line of output (3) as you would expect.
--
D. Richard Hipp <[EMAIL PROTECTED]>
ength allocation is
roughly the same thing.
--
D. Richard Hipp <[EMAIL PROTECTED]>
e taken on unused databases.
--
D. Richard Hipp <[EMAIL PROTECTED]>
new features go into version 3.
--
D. Richard Hipp <[EMAIL PROTECTED]>
er.
--
D. Richard Hipp <[EMAIL PROTECTED]>
crosscompiler -o libsqlite.a *.c
--
D. Richard Hipp <[EMAIL PROTECTED]>
f the isNT() function?
--
D. Richard Hipp <[EMAIL PROTECTED]>
On Sun, 2005-06-19 at 22:29 -0400, D. Richard Hipp wrote:
> Essentially the same talk will likely be repeated
> at other venues. I will keep you posted.
If you cannot make it to OSCON or AUUG, the next
opportunity to hear the SQLite Internals tutorial
will be at the 12th Annual Tcl/Tk Conf
l section 2.1
for details.
--
D. Richard Hipp <[EMAIL PROTECTED]>
force the value to be inserted as string?
>
Make the declared datatype of the column TEXT.
Example:
CREATE TABLE t1(a, b INTEGER, c REAL, d TEXT);
INSERT INTO t1 VALUES(9.0,9.0,9.0,9.0);
SELECCT * FROM t1;
Results in:
9|9|9|9.0
--
D. Richard Hipp <[EMAIL PROTECTED]>
On Mon, 2005-06-20 at 15:03 +0300, Amir Hadar wrote:
> Hi
>
> I encountered an assert in function moveToRoot in file btree.c.
>
Not much to go on. Can you tell me what you were doing
or how to reproduce the problem?
--
D. Richard Hipp <[EMAIL PROTECTED]>
ke
a case that there is sufficient interest, I'm sure
something could be worked out.
--
D. Richard Hipp <[EMAIL PROTECTED]>
yet been accepted, though.
--
D. Richard Hipp <[EMAIL PROTECTED]>
On Sun, 2005-06-19 at 20:15 -0400, Paul G wrote:
> i have other obligations which will prevent me from being able to travel to
> oregon at that time.
>
Essentially the same talk will likely be repeated
at other venues. I will keep you posted.
--
D. Richard Hipp <[EMAIL PROTECTED]>
On Sun, 2005-06-19 at 21:29 +0200, Marcel Strittmatter wrote:
> > http://conferences.oreillynet.com/oscon2005/
> >
>
> http://conferences.oreillynet.com/os2005/ is the correct url, isn't it?
>
Yes it is. Sorry for the typo.
--
D. Richard Hipp <[EMAIL PROTECTED]>
products
* Anyone who is just curious about how SQL database engines
work.
For additional information visit
http://conferences.oreillynet.com/oscon2005/
I hope to see many of you there.
--
D. Richard Hipp <[EMAIL PROTECTED]>
in than needed, or something
> else?
>
Why not use a smaller page size but map multiple pages at a time?
--
D. Richard Hipp <[EMAIL PROTECTED]>
nt.
See also http://www.sqlite.org/copyright.html
--
D. Richard Hipp <[EMAIL PROTECTED]>
works for memmap() would also continue to work
for the old read()/write() method. So presumably you could create
a new pragma that lets the programmer decide at run-time which method
to use. Use memmap() for faster reads and less memory usage or
use read/write for better concurrency.
--
D. Richar
uot; on the
compiler
I do use (GCC) over the years. In every case, the cause has turned out
to
be a defective memory SIMM.
Memtest86 is your friend. http://www.memtest86.com/
--
D. Richard Hipp <[EMAIL PROTECTED]>
write just the one page where the value
is being inserted.
--
D. Richard Hipp <[EMAIL PROTECTED]>
s of
posix locks and the very specific brokenness of posix locks on RH9.
There aren't really any other options here.
> That
> is, should the FAQ be updated to indicate that only the thread that calls
> sqlite3_open is allowed to use the structure?
>
The documentation needs to be upda
(which I do not use so am
unlikely to ever notice.) Fixed now.
--
D. Richard Hipp <[EMAIL PROTECTED]>
etrying the failed query. Seems to work...
>
There should never be such a thing as a transient
SQLITE_CORRUPT error. You should only get an SQLITE_CORRUPT
error if the database file is truly and permanently
corrupted.
--
D. Richard Hipp <[EMAIL PROTECTED]>
s undefined, it is assumed to be zero. No changes
are needed to make this work.
--
D. Richard Hipp <[EMAIL PROTECTED]>
On Wed, 2005-06-15 at 12:43 -0400, Christopher R. Palmer wrote:
> I've temporarily placed the trace file at
>
> http://67.19.10.2/~crpalmer/sqlite-traking.gz
>
The file is named sqlite-trace-locking.gz. And it
did not containing the debugging information I need.
Perhaps you didn't compile with
me wrong...
--
D. Richard Hipp <[EMAIL PROTECTED]>
. Richard Hipp <[EMAIL PROTECTED]>
, parse.y.
If your (broken) compiler insists on having parse.y present, then why
not just download it from the website.
http://www.sqlite.org/cvstrac/getfile/sqlite/src/parse.y
--
D. Richard Hipp <[EMAIL PROTECTED]>
ents. It then returns a pointer to the beginning of the
second statement so that you can process them all in a loop. But
your code appears to omit this loop and thus is processing only
the first one. It never reaches the SELECT.
--
D. Richard Hipp <[EMAIL PROTECTED]>
e from moral rights. That is
too say, in Canada moral rights apply to works of art and literature
but not to computer programs. What is the situation in Europe? Are
you sure that moral rights apply to software there?
--
D. Richard Hipp <[EMAIL PROTECTED]>
ated to the public
domain seems unimportant.
--
D. Richard Hipp <[EMAIL PROTECTED]>
etId=203 AND SectId=4
UNION ALL
SELECT * FROM PointFeature WHERE DsetId=203 AND SectId=400;
--
D. Richard Hipp <[EMAIL PROTECTED]>
parate transaction. 5
million inserts per days is 58 transactions per second - which
is aggressive. It will work better to batch your inserts
in order to keep the number of transactions below 1 or
2 per second.
--
D. Richard Hipp <[EMAIL PROTECTED]>
e
the last time sqlit3_prepare was run. All you have
to do is rerun sqlite3_prepare. Closing and opening
the database first is not necessary.
--
D. Richard Hipp <[EMAIL PROTECTED]>
r to properly
integrate SSE. It is those (relatively minor) adjustments that
you are seeing on the SQLite website.
--
D. Richard Hipp <[EMAIL PROTECTED]>
is confusing
the optimizer somehow. (I'm still looking why this is.) SQL
(not just SQLite but the SQL language in general) really wants
you to use single quotes. If you change "NL" to 'NL' the query
will go very quickly, even without creating any indices.
--
D. Richard Hipp <[EMAIL PROTECTED]>
On Sun, 2005-05-22 at 21:12 +0200, Ludvig Strigeus wrote:
> If I corrupt my database in certain ways, I can make Sqlite crash. Is
> this by design, or is it a bug?
>
This is a bug, though not a high-priority one.
--
D. Richard Hipp <[EMAIL PROTECTED]>
On Sun, 2005-05-22 at 20:35 +0200, Ludvig Strigeus wrote:
> How do I run the unit tests in Linux?
>
> I've managed to build "tclsqlite3", but where do I go from there?
>
make testfixture
./testfixture ../sqlite/test/all.test
Or simply
make fulltest
--
D
ange to pass a full regression test, I will likely
check it in.
--
D. Richard Hipp <[EMAIL PROTECTED]>
> reveals other developers having the same issue.
>
> Is it just me or is 'full_column_names' still broken?
>
Set short_column_names=off first, then full_column_names
will work as it did before.
--
D. Richard Hipp <[EMAIL PROTECTED]>
ver the locking bytes.
--
D. Richard Hipp <[EMAIL PROTECTED]>
an probably also think of schemes where you store either
the SELECT results or the UPDATE statements in memory. Note
that technique (2) above works by moving the entire result set
into memory for you. Avoid solution (2) if your result set is
exceedingly large.
--
D. Richard Hipp <[EMAIL PROTECTED]>
s a link to this (called "timeline") on the top-right
of the webpage header. The blue dots are CVS check-ins. If you
do not care to see the other stuff (ticket and wiki changes)
there are options at the bottom of the page to omit those things.
--
D. Richard Hipp <[EMAIL PROTECTED]>
of all the
details for you.
--
D. Richard Hipp <[EMAIL PROTECTED]>
orry about being
atomic and durable across power failures, they can get away
with a lot of things that a disk database cannot.
--
D. Richard Hipp <[EMAIL PROTECTED]>
MMIT faster than this and still
be ACID.) If you fail to enclose your INSERTs in a single
BEGIN...COMMIT then separate BEGIN...COMMITs are automatically
added around each INSERT, which really slows things down.
--
D. Richard Hipp <[EMAIL PROTECTED]>
On Tue, 2005-05-03 at 18:25 -0400, Christopher Petrilli wrote:
> On 5/3/05, D. Richard Hipp <[EMAIL PROTECTED]> wrote:
> > On Tue, 2005-05-03 at 17:52 -0400, Christopher Petrilli wrote:
> > > I'm wondering if the availability of 128-bit numbers has ever been
> >
On Tue, 2005-05-03 at 18:55 -0400, Tom Shaw wrote:
> At 6:09 PM -0400 5/3/05, D. Richard Hipp wrote:
> >On Tue, 2005-05-03 at 17:45 -0400, Tom Shaw wrote:
> >> Hi.
> >>
> >> I was using SQLite with PHP 5 (MacOSX) and due to some issues the php
> >>
n that PHP
is somehow holding the database open (and locked) after it
times out. Try shutting down and restarting your webserver
and see if that doesn't clear the problem.
--
D. Richard Hipp <[EMAIL PROTECTED]>
On Tue, 2005-05-03 at 15:09 +0100, Drew, Stephen wrote:
> This is not the case for the version of SQLite 3.2.1 I am using. Is
> the documentation incorrect, or am I doing anything wrong?
>
Documentation is wrong. See http://www.sqlite.org/cvstrac/wiki?
p=ColumnNames
--
D. Ric
d into the
standard release in the near future.
--
D. Richard Hipp <[EMAIL PROTECTED]>
e copying it. The COMMIT statement release
the lock after you have finished making the copy.
--
D. Richard Hipp <[EMAIL PROTECTED]>
On Sun, 2005-05-01 at 16:19 +1000, Dennis Volodomanov wrote:
> Hello all,
>
> I'm getting an assert(pCur->isValid) i
Please provide the schema and an SQL statement that makes this
happen.
--
D. Richard Hipp <[EMAIL PROTECTED]>
On Sat, 2005-04-30 at 13:51 +0200, VÃclav Haisman wrote:
> Hi, for me to be able to use SQLite3 with C++ sources that use -ansi
> -pedantic with GCC I need following tiny patch:
>
Wouldn't it be easier to *not* compile with -ansi -pedantic?
--
D. Richard Hipp <[EMAIL PROTECTED]>
. It is part of
the core library and has been for time out of mind. I do not
have any ideas about what could be going wrong.
--
D. Richard Hipp <[EMAIL PROTECTED]>
t content without making unnecessary copies or conversions.
The technique above runs much faster faster. And, the latest Tcl
interface caches prepared statements, making it faster still if
the same statement is run multiple times.
--
D. Richard Hipp <[EMAIL PROTECTED]>
onfigured as a cross-compiler. I highly
recommend that solution as it seems to work much better than
trying to compile things under windows.
--
D. Richard Hipp <[EMAIL PROTECTED]>
On Wed, 2005-04-27 at 14:40 -0500, Alexander Thomason wrote:
> the database is located at http://www.ebrats.org/site/forum/board2.db
That URL does not work for me. Are you sure that is were the
database is suppose to be?
--
D. Richard Hipp <[EMAIL PROTECTED]>
On Wed, 2005-04-27 at 13:19 +, Tiago Dionizio wrote:
> On 4/27/05, D. Richard Hipp <[EMAIL PROTECTED]> wrote:
> > I have closed all these tickets now. See wiki page
> > http://www.sqlite.org/cvstrac/wiki?p=ColumnNames
>
> Isn't the pragma long_column_names suppo
tion work for you?
--
D. Richard Hipp <[EMAIL PROTECTED]>
w.sqlite.org/cvstrac/wiki?p=ColumnNames
--
D. Richard Hipp <[EMAIL PROTECTED]>
On Tue, 2005-04-26 at 17:49 +0100, Christian Smith wrote:
> Just created ticket #1224 to remove config.h from build, but there appears
> to be no way to attach a patch to the ticket itself. Have I missed
> something?
>
There is an "[Attach]" hyperlink at the top of the
This is a test of the mailing list software.
Please ignore.
--
D. Richard Hipp <[EMAIL PROTECTED]>
On Tue, 2005-04-26 at 11:56 +0930, Jake Skinner wrote:
> Does anyone know if there is a maximum number of precompiled queries you
> can attach to a database?
There is no arbitrary limit. You can have as many precompiled
statements as you have memory to store them in.
--
D. Richard Hipp &
a command-line shell
and type:
select sqlite_version();
That will tell you right away what version is being used.
I do not know, but I suspect they did their final builds
of Tiger with 3.1.5.
--
D. Richard Hipp <[EMAIL PROTECTED]>
fied or documented before, I'd be
willing to do it since any code that breaks would have been
depending on undefined behavior to begin with.
Does the SQL standard have anything to say about this?
Does anybody know?
--
D. Richard Hipp <[EMAIL PROTECTED]>
d run in O(logN) time. Without those changes, the
execution time should be O(N^2).
--
D. Richard Hipp <[EMAIL PROTECTED]>
get an error? Please explain the problem
> > in more detail.
>
> Sorry I was unclear.
>
> I open the database, do some inserts(). idle for 3-4 hours without
> closing the handle and then do some selects it will give me that error.
>
Sounds to me like some other process
quot;order by" support for any language you want
with the sqlite3_create_collation() API.
> - it is not suppports functions like upper/lower (with hun chars)
Use the sqlite3_create_function() API to replace the built-in
upper/lower functions with versions that work on your language.
--
D. Richard Hipp <[EMAIL PROTECTED]>
was another method to return the same error that
> finalize() returns (without destroying the statement).
Use sqlite3_reset() instead of sqlite3_finalize().
--
D. Richard Hipp <[EMAIL PROTECTED]>
quot;).
What do you mean "holding the DB open"? If you you do not send
it any queries, you can you get an error? Please explain the problem
in more detail.
--
D. Richard Hipp <[EMAIL PROTECTED]>
On Thu, 2005-04-21 at 18:18 -0700, Vladimir Vukicevic wrote:
> On 4/21/05, D. Richard Hipp <[EMAIL PROTECTED]> wrote:
> > So when the schema changes, it is not a matter of changes
> > a few bits and pieces of the VM. The VM must be completely
> > reconstructed. Its a
complete do-over.
--
D. Richard Hipp <[EMAIL PROTECTED]>
e...
>
COLUMN became a reserved word when the ALTER TABLE ADD COLUMN statement
was added. You can work around the problem by putting the name of
column "column" in quotes.
--
D. Richard Hipp <[EMAIL PROTECTED]>
> I do not unfortunately have any place to host this work. There may be
> somebody on this list who knows of a place where this can be published
> for any members who may with make use of it.
>
http://www.sqlite.org/contrib
--
D. Richard Hipp <[EMAIL PROTECTED]>
letely. So SQLite
might end up using more indices than tables. But the number
of indices used will never exceed the number of tables named
in the query.
--
D. Richard Hipp <[EMAIL PROTECTED]>
is available for windows and comes
preinstalled on everything else. Or you can just look at
the patch and fix the problem in your local copy. (The bug
fix involves transposing two characters on a single line.)
I do not have a schedule for the release of the next version
of SQLite.
--
D. Richard Hipp <[EMAIL PROTECTED]>
ly your friend here, at least
> for the time being.
>
sqlite3_mprintf() with the %q parameter is perhaps more friendly
still.
--
D. Richard Hipp <[EMAIL PROTECTED]>
an ATTACH
statement - that would make no sense. Hence, bound parameters are
not allowed in that context.
--
D. Richard Hipp <[EMAIL PROTECTED]>
901 - 1000 of 1539 matches
Mail list logo