- users-list - monetdb.org

[MonetDB-users] MonetDB XQuery Bulk Load / Bulkload
by Alexander Koglin 07 Feb '10

07 Feb '10

I have currently installed a version of MonetDB XQuery for Linux and want to bulkload data like this: http://monetdb.cwi.nl/XQuery/Documentation/Bulk-Loading-a-Collection.html#B… For that purpose, I used the following command inside the Client: for $d in doc("/monetdbfiles.xml")//doc return pf:add-doc(fn:concat($d/@path,$d/@name), fn:string($d/@name), "abap10k", 0) <> It references the following, definitely existing XML file: <dir> <doc path="/home/epic/tarari_svn/benchmarks/abap/ABAP/" name="ACL_AUNIT_RESULT==============CCDEF"> <doc path="/home/epic/tarari_svn/benchmarks/abap/ABAP/" name="ACL_AUNIT_RESULT==============CCIMP"> <doc path="/home/epic/tarari_svn/benchmarks/abap/ABAP/" name="ACL_AUNIT_RESULT==============CCMAC"> <doc path="/home/epic/tarari_svn/benchmarks/abap/ABAP/" name="ACL_AUNIT_RESULT==============CI"> <doc path="/home/epic/tarari_svn/benchmarks/abap/ABAP/" name="ACL_AUNIT_RESULT==============CM001"> <doc path="/home/epic/tarari_svn/benchmarks/abap/ABAP/" name="ACL_AUNIT_RESULT==============CM002"> <doc path="/home/epic/tarari_svn/benchmarks/abap/ABAP/" name="ACL_AUNIT_RESULT==============CM003"> <doc path="/home/epic/tarari_svn/benchmarks/abap/ABAP/" name="ACL_AUNIT_RESULT==============CM004"> <doc path="/home/epic/tarari_svn/benchmarks/abap/ABAP/" name="ACL_AUNIT_RESULT==============CM005"> </dir> But unfortunately an error is thrown (I also tried it unter Windows and it was the same): MAPI = root@localhost:50000 QUERY = for $d in doc("/monetdbfiles.xml")//doc ERROR = !ERROR: err:FODC0002, Error retrieving resource (no such document "/monetdbfiles.xml"). What could be the reason for that?

1 1

[MonetDB-users] authorize! what's wrong?
by korisk 07 Feb '10

07 Feb '10

Hi! I've installed monetdb from sources. ./MonetDB5-server-5.16.4.tar.bz2 ./MonetDB-client-1.34.4.tar.bz2 ./MonetDB-SQL-2.34.4.tar.bz2 merovingan started database test created but mclient asks for password! and i don't know which one. Manual says that monetdb uses system authorization, but mclient answers: korisk@pс64:~$ mclient -l sql -d test user(korisk): password: InvalidCredentialsException:checkCredentials:invalid credentials for user 'korisk' i use Ubuntu 9.10 x86_64 What's wrong? Thank you.

2 1

[MonetDB-users] authorization! what's wrong?
by korisk 07 Feb '10

07 Feb '10

Hi! I've installed monetdb from sources. ./MonetDB5-server-5.16.4.tar.bz2 ./MonetDB-client-1.34.4.tar.bz2 ./MonetDB-SQL-2.34.4.tar.bz2 merovingan started database test created but mclient asks for password! and i don't know which one. Manual says that monetdb uses system authorization, but mclient answers: korisk@pс64:~$ mclient -l sql -d test user(korisk): password: InvalidCredentialsException:checkCredentials:invalid credentials for user 'korisk' i use Ubuntu 9.10 x86_64 What's wrong? Thank you.

2 2

[MonetDB-users] connecting streams to Mapi.
by Guillaume Yziquel 05 Feb '10

05 Feb '10

Hi. I'm investigating how to connect custom streams to a Mapi structure. I've been reading, for instance the start_mal_server() function of monetdb5-server-5.16.2/src/tools/embeddedclient.c.in. I now sort of understand how the stream library works (though not fully), and I'm trying to understand how the streams are effectively used by the client of the Mapi struct. I've therefore looking at the start_mal_server() function, and here's a piece of code whose semantics I do not understand: > in = ((stream **) arg)[0]; > out = ((stream **) arg)[1]; > free(arg); > snprintf(buf, sizeof(buf), > "in := \"" PTRFMT "\":streams;\n" > "out := \"" PTRFMT "\":streams;\n" > "mapi.malclient(in, out);\n", PTRFMTCAST in, PTRFMTCAST out); > callString(mal_clients, buf, 0); I'm specifically interested in the 'in' stream since it is the one I would like to write to. Could someone enlighten me here? All the best, -- Guillaume Yziquel http://yziquel.homelinux.org/

2 4

[MonetDB-users] Feb2010 64 bit starts and immediately stops on Windows
by Alex Bo. 05 Feb '10

05 Feb '10

Hi dear MonetDB people, I built MonetDB 64 bit on Windows from the Feb2010 branch today. Everything ran fine, but when I start the server, after seeing the normal output, it immediately dissapears. Here's the output: # MonetDB server v5.18.0, based on kernel v1.36.0 # Serving database 'demo', using 1 thread # Compiled for i686-pc-win32/64bit with 32bit OIDs dynamically linked # Found 1.999 GiB available main-memory. # Copyright (c) 1993-July 2008 CWI. # Copyright (c) August 2008-2010 MonetDB B.V., all rights reserved # Visit http://monetdb.cwi.nl/ for further information #warning: please don't forget to set your vault key! #(see C:\MonetCVS_base_Feb10\MonetCVS64_cleanBuild\Install\etc\monetdb5.conf) # Listening for connection requests on mapi:monetdb://127.0.0.1:50000/ # MonetDB/SQL module v2.36.0 loaded Building MonetDB 32 bit works just fine. Thanks. -- View this message in context: http://old.nabble.com/Feb2010-64-bit-starts-and-immediately-stops-on-Window… Sent from the monetdb-users mailing list archive at Nabble.com.

2 1

[MonetDB-users] MCinitClient() and user oid.
by Guillaume Yziquel 04 Feb '10

04 Feb '10

Hi. I've been trying to get access to the Client structures on the server side. I've therefore been looking at MCinitClient(). What I do not understand is how do you know which oid for the user you should pick, in order to pass this oid as an argument to MCinitClient. > http://yziquel.homelinux.org/topos/api/ocaml-monetdb5/MonetDB5.MAL.Client.h… > > Client MCinitClient(oid user, bstream *fin, stream *fout) Moreover, it seems to me that in an embedded setting, it doesn't seem to be a smart idea to use MCinitClient. It would perhaps be better to grab the already existing Client structure for the admin. Probably with MCfindClient. > http://yziquel.homelinux.org/topos/api/ocaml-monetdb5/MonetDB5.MAL.Client.h… > > Client MCfindClient(oid user, str scen, bstream *fin, stream *fout) fin and fout seem optional arguments (meaning you can pass NULL for sensible defaults). But as before, the oid of the user, and the sceanrio string are arguments whose proposed values are unclear to me. How do you get the admin's Client struct? All the best, -- Guillaume Yziquel http://yziquel.homelinux.org/

3 7

[MonetDB-users] copy into with 'float' types crashes
by Alex Bo. 04 Feb '10

04 Feb '10

Hi, I am encountering something very strange. (Windows - NovSP1 from CVS) On some machines I encountered crashes in Monet when doing a copy into on tables with a float column. After trying to investigate, I narrowed it down to this: Run: 1. CREATE TABLE test (f1 int, f2 float); 2. COPY INTO test from 'filename.txt' using delimiters ',','\n','"' null as ''; the text file contains: 1,1 2,2 This sequence exactly caused a server crash (without any message) on some machines, while on others it didn't. The machines it happened on were: 1. 32 bit XP with 3 GB RAM. 2. 32 bit Vista with 4 GB RAM. In both cases, changing the column type from float to real worked. Could it be that float, being a 64 bit floating point, might cause crashes on 32 bit machines? Is there anything I can to help you get a clearer picture of this? Thanks. -- View this message in context: http://old.nabble.com/copy-into-with-%27float%27-types-crashes-tp27438598p2… Sent from the monetdb-users mailing list archive at Nabble.com.

3 4

[MonetDB-users] The maximum used memory option
by Stefan de Konink 04 Feb '10

04 Feb '10

It was discussed before that there should exist an option to limit the amount of 'detected' memory to a certain value. Is this option available, if so, what is its syntax? The other way around was to use ulimit on the process, does this mean MonetDB will get a sigkill if it grows out of boundaries or will it just use that value as upper memory limit? Stefan

2 1

[MonetDB-users] Problem using win 32 odbc driver
by klaas＠maxml.nl 04 Feb '10

04 Feb '10

Good morning, I'm using MonetDB v5.16.2 on windows xp 32. Connecting to the database with Java clients works fine, odbc (version 10.1.12) is not. Table DDL: CREATE TABLE _r_language( id smallint NOT NULL AUTO_INCREMENT, abbrev varchar(3) NOT NULL, name varchar(50) NOT NULL, primary key (id) ); For example in php: <?php $conn = odbc_connect('DSN=MonetDB;HOST=localhost;PORT=50000', 'monetdb', 'monetdb') or die('unable to connect'); $result = odbc_exec($conn, 'select * from _r_language;'); odbc_result_all($result); ?> results in: ? ?? ??? 1 English EN 2 German DE 3 French FR 4 Dutch NL 5 Polish PL The fieldnames are not returned. Using the same ODBC driver in MS Access, the table can be linked, column_names are shown, but retrieving data generates error message: ODBC - call failed [MonetDb][ODBC Driver 1.0]TODO: column names of level >= 3 (#0) What can you recommend? Thanks for your help, regards, Klaas

1 0

[MonetDB-users] Types oid, size_t.
by Guillaume Yziquel 03 Feb '10

03 Feb '10

Hi. I've been wanting to wrap up oids into OCaml, and I'm wondering a few things about the following statement in gdk.mx: > @item oid: > Unique @strong{long int} values uses as object identifier. Highest bit cleared always. > Thus, oids-s are 31-bit numbers on 32-bit systems, and 63-bit numbers on 64-bit systems. Moreover, the oid type is typedefed in the following way: > #ifdef MONET_OID32 > #define SIZEOF_OID SIZEOF_INT > typedef unsigned int oid; > #else > #define SIZEOF_OID SIZEOF_SIZE_T > typedef size_t oid; > #endif I understood that MONET_OID32 was the exception rather than the rule, so I'm assuming, for now, that oid is size_t. But then, I'm wondering why the highest bit is always cleared. Moreover, OCaml has a way to deal with their integers which is rather peculiar. They are long ints, but the lowest bit has a peculiar meaning. If the lowest bit is 0, then this long int is in fact a pointer to an OCaml value in the OCaml heap. If the lowest bit is 1, then shift the value one bit, and you have the Ocaml integer. This scheme takes advantage of the alignement of OCaml values in the OCaml heap. So ocaml integers are 31-bit on 32-bit systems, and 63-bit on 64 bit systems. The difference is that the lowest bit is set, whereas it's the highest bit that is cleared for oids. Therefore, to convert oids to ocaml ints and conversely, I wrote: > #define ocamlmonetdb5_Val_oid(x) ((value) ((((oid) x) << 1) + 1)) > #define ocamlmonetdb5_Oid_val(x) (((oid) x) >> 1) Two questions: -1- This seems to work, but I'd appreciate if you could double-check this. Just to check if I understood correctly the bit layout of oids. -2- I'm wondering why the highest bit of oids is always cleared. This doesn't seem to be implied by the fact that oids are typedefed size_t. So what's the rationale? All the best, -- Guillaume Yziquel http://yziquel.homelinux.org/

1 0