OpenEuphoria

1. SQLite wrappers

• Posted by Tone Škoda <tskoda at email.si> Aug 23, 2005
• 580 views

I want to convert this demo to Euphoria, because I need fast insert:
http://anchor.homelinux.org/SQLiteTuning#head-0dd2e45aa0ac71a09ca7860dc4bc435f5e8bdbc5

To do this I need to make wrappers for many functions which have reference here:
http://sqlite.org/capi3ref.html


Before I start doing it, did by any chance someone already do this work so I
don't have to do it?

3. Re: SQLite wrappers

• Posted by Tone Škoda <tskoda at email.si> Aug 24, 2005
• 554 views

jacques deschênes wrote:

> Search the archive for eusqlite


i already did that, those wrappers are incomplete. and sqlite_exec() doesn't
work, i don't know why.
anyway i am going to do the wrappers for those additional functions and since c
pointers are still blurry to me i made me this tutorial to clear them up:
http://www10.brinkster.com/tskoda/C_pointers_explanation.htm
maybe someone will find it helpful.

5. Re: SQLite wrappers

• Posted by DB James <larch at adelphia.net> Aug 24, 2005
• 566 views

Tone Škoda wrote:
<SNIP>
> anyway i am going to do the wrappers for those additional functions and since
> c pointers
> are still blurry to me i made me this tutorial to clear them up:
> <a
> href="http://www10.brinkster.com/tskoda/C_pointers_explanation.htm">http://www10.brinkster.com/tskoda/C_pointers_explanation.htm</a>
> maybe someone will find it helpful.
> 

Hi Tone,

I glanced over your self-explanation of C pointers are, how they work, and why
they are confusing.  I agree there is something elusive about them and I've seen
C expressions that are as weird looking as the stranger propositions in
metaphysics: pointer variables to arrays of pointers, etc.

I wonder if, in the beginning of C, there had been "val", "adr", and "ind"
(indirect) keywords -- would that have helped?  Assuming a language called maybe
Explicit C, all references to variables had these terms, then probably there
would not be this confusion of memory-location and
value-contained-in-that-location.

if bar is in memory address 10 and that contains the number 20 and the value
stored in memory address 20 is 30, then

val foo = adr bar --foo now equals 10
val foo = val bar --foo now equals 20
val foo = ind bar --foo now equals 30

To load an offset, i.e. a memory location:
val count+=1
val foo = val (val bar + val count)

To load a value at an offset:
val count+=1
val foo = ind (val bar + val count)

To call a routine that expects a number:
MyProcedure(val normaJean)

To call a routine that expects an address:
x=MyFunction(adr marilyn)

So, all left-items are always just numbers, and in practice the "val" could be
left off.  What use there is in *var=whatever, I don't know.  It is still just a
number, like everything in computers.  My impression is that it is "typing" gone
wild, where not only the byte-length of a variable is declared, but the
interpretation of the numbers as well.

Just some idle musings...

--Quark

7. Re: SQLite wrappers

• Posted by don cole <doncole at pacbell.net> Aug 25, 2005
• 565 views

Tone ,

I checked out you tutorial and found it to be very clear and precise and 

interesting.

Although I tickered the C Borland tutorial, I never got into it too much

because I thought C was too complicated (which is why I chose Euphoria as my

primary language).

However should I deleve into C further I will keep your tutorial in mind.

Also from your tutorial I can see how C and Euphoria are closely related.


Kat,

All the lines in your messages end in =

then on the next line one character follows and then next line.

Making it very hard on the eyes too read.

I don't know what's up with that.

Just thought I'd let you know.

Don Cole,
SF

9. Re: SQLite wrappers

• Posted by Chris Burch <chriscrylex at aol.com> Aug 29, 2005
• 584 views
• Last edited Aug 30, 2005

Hi

Just got back from hols in Ballamory.

I used eusqlite as it stands atm, and created this. Note sqlite_get_table is a
wrapper for sqlite_exec, and as yet I haven't needed anything more complex than
what the library provides, but looking at the benchmarks, it may be that sqlite
prepared statements may be quite a bit faster than throwing sql statements at the
engine one at a time - will look into this.

Note you can use begin / end / commit transactions with what is there already,
and as it stands the eusqlite libs are very usable.

Also I'm not sure how much  eu, or my code, is contributing to the slowdown (any
contributions from the speed meisters out there gratefullt accepted!)

--a close 'equivalent' to the c program at
--
http://anchor.homelinux.org/SQLiteTuning#head-0dd2e45aa0ac71a09ca7860dc4bc435f5e8bdbc5
-- sqlite_get_table is a wrapper for sql_exec, and we can't prepare statements
(yet)
-- this is quite a bit slower than the compiled c versions

without warning

include eusqlite3.ew
include get.e

object VOID
VOID = {}


------------------------------------------------------------------------
function eusql_begin(integer db)
------------------------------------------------------------------------
sequence sqlite_data

sqlite_data = sqlite_get_table(db, "BEGIN")     --shorthand for begin
transaction

if sqlite_last_err_no != SQLITE_OK then
printf(1, "couldn't begin transaction:  %s\n", {sqlite_last_err_desc})
        return 0
end if

return 1
end function

------------------------------------------------------------------------
function eusql_commit(integer db)
------------------------------------------------------------------------
sequence sqlite_data

sqlite_data = sqlite_get_table(db, "END")     --shorthand for end transaction,
and it also commits the transaction

if sqlite_last_err_no != SQLITE_OK then
printf(1, "couldn't comit transaction:  %s\n", {sqlite_last_err_desc})
        return 0
end if

return 1
end function

------------------------------------------------------------------------
function elapsed(atom start_time, atom end_time)
------------------------------------------------------------------------
return end_time - start_time
end function


------------------------------------------------------------------------
--main
------------------------------------------------------------------------

--variables
integer db, N, xact_size, n_this_xact
sequence cmd, temp_seq, sql_data, sql_cmd
atom START, FINISH, x, y, z

cmd = command_line()
if length(cmd) < 3 then
        printf(1,"\nUsage:   %s  <N>  <X>\n\n", {cmd[2] })    
        puts(1,"\tInsert <N> rows into a table of an SQLite database\n")    
        puts(1,"\tusing transaction sizes of <X>.\n")    
        puts(1,"\tThe table has four columns of numeric data:\n")    
        puts(1,"\t  field_1 integer (primary key)\n")    
        puts(1,"\t  field_2 float\n")    
        puts(1,"\t  field_3 float\n")    
        puts(1,"\t  field_4 float\n")    
        puts(1,"\tThe integer field will have values 1..<N> while the\n")    
        puts(1,"\tdouble precision values are random on [-50.0, 50.0]\n")    
        ? 1/0    
end if

--interpret command line
temp_seq = value(cmd[3])
N = temp_seq[2] 

temp_seq = value(cmd[4])
xact_size = temp_seq[2]

--database shouldnt exist before start, so
if platform() = LINUX then
	system("rm bench.db", 0)
else
	system("del bench.db", 0)
end if


puts(1, "Running.......\n")
START = time()

--we do this a slightly different way to the C example,
--but the end result is (hopefully) the same, or very similar.
--sqlite_get_table is a wrapper for sqlit3_exec, the returns sre the same

db = sqlite_open("bench.db", 0)

sql_data = sqlite_get_table(db, "PRAGMA synchronous = OFF;")

VOID = eusql_begin(db)
sql_data = sqlite_get_table(db, "create table table_name(field_1 integer primary
key, " &
                                                     "field_2 float,  " &
                                                     "field_3 float,  " &
                                                     "field_4 float)")

--we can't prepare statements - yet, maybe one day if it ever looks like it may
be useful
--so we will have to create the statements one at a time (sorry)
n_this_xact = 0
for i = 0 to N do
        x = 50 - (rand(300) / 3.141592654)        
        y = 50 - (rand(300) / 3.141592654)        
        z = 50 - (rand(300) / 3.141592654)
        
sql_cmd = sprintf("insert into table_name (field_2, field_3, field_4)
        values (%f, %f, %f)", {x, y, z})
        --note, field_1 autoincrements

        sql_data = sqlite_get_table(db, sql_cmd)
        if sqlite_last_err_no != SQLITE_OK then
                printf(1, "Error :  %s\n", {sqlite_last_err_desc})
                VOID = wait_key()
                abort(0)
        end if

        n_this_xact += 1

        if n_this_xact >= xact_size then
                n_this_xact = 0
                VOID = eusql_commit(db)
                VOID = eusql_begin(db)
        end if
end for
VOID = eusql_commit(db)
sqlite_close(db)

FINISH = time()

printf(1," %d inserts to %s in %.3f s = %.2f inserts/s\n", {N, "bench.db",
elapsed(START, FINISH), N/elapsed(START, FINISH)  })

VOID = wait_key()


--Sample Results from c program
--Pentium4 3.0 GHz, 1 GB RAM, IDE disk drive, ext3, RHEL v3, Linux kernel
2.4.21:
--
--./sqlite_insert 100000 20000
--100000 inserts to bench.db in 0.829 s = 120626.53 inserts/s
--Opteron 2.4Ghz, 4GB RAM, Seagate 10k SCSI disk w/ XFS noatime, 2.4.27 Kernel: 
--
--ponte [/local/scratch/kcroft] ./sqlite_insert 200000 60000
--200000 inserts to a.db in 0.646 s = 309836.37 inserts/s
--last edited 2005-05-20 16:56:18 by adsl-69-234-50-233

--Sample results, from eu program
--exw sqlite_insert.exw 100000 20000
--100000 inserts to bench.db in 13.230 s = 7558.58 inserts/s
--Dell inspiron 2200 1.3Ghz 256Mb ram win xp
--200000 inserts to bench.db in 32.660 s = 6123.70 inserts/s

--Dell 2.4Ghz, 512Mb
-- 100000 inserts to bench.db in 6.718 s = 14885.38 inserts/s

Chris

http://members.aol.com/chriscrylex/euphoria.htm
http://uboard.proboards32.com/
http://members.aol.com/chriscrylex/EUSQLite/eusql.html