- users-list - monetdb.org

Drop a local temporary table within a function
by Luis Larios 14 Jul '17

14 Jul '17

Hi, all, Consider the following scenario. I have a bunch of user defined functions, f_i, that create, populate, and use a local temporary table with the same name "res": create function f_i() returns table( ... ) begin # Result stats table. create local temporary table res( param1 int, param2 int ); /* Populate res */ /* Use res in some queries */ return table( /* Query involving res */ ); end; My first question is: Does "create local temporary table x" create a table that is disposed/dropped as soon as the function where it was declared finishes execution? I assumed this was the case, but I'm now skeptical as we are running into all sorts of concurrency issues that point to the "res" table not being cleared out properly from memory. This takes me to the second question: How do I drop a temporary table if such table exists? The reason for this is to prevent "table create" exceptions (which we are currently running into). I'd like to achieve something along these lines: create function f_i() returns table( ... ) begin # Drop local temporary table if such table exists. drop local temporary table res; <-------- How do I do this in MonetDB? # Result stats table. create local temporary table res( param1 int, param2 int ); /* Populate res */ /* Use res in some queries */ return table( /* Query involving res */ ); end; Where, as you can see, I want to explicitly drop the local temporary table "res" if such exists. If I leave the "local temporary" markers in the drop statement, MonetDB complains that it found unexpected "LOCAL" or "TEMPORARY". I wouldn't like to remove these markers because a statement like: drop table res; will drop any table in the default function schema with that name. Any help and hints are well appreciated! Thanks for the help in advance ~ Luis Angel The information transmitted, including any attachments, is intended only for the individual or entity to which it is addressed, and may contain confidential and/or privileged information. Any review, retransmission, dissemination or other use of, or taking of any action in reliance upon, this information by individuals or entities other than the intended recipient is prohibited, and all liability arising therefrom is disclaimed. If you have received this communication in error, please delete the information from any computer and notify the sender.

3 2

Jul2015-SP4 vs. Dec2016-SP5
by Anderson, David B 13 Jul '17

13 Jul '17

All, I know I've skipped multiple releases so this is a difficult question to answer, but should I expect the Dec2016-SP5 version of MonetDB to be faster than/equal to the Jul2015-SP4 version? I have a parallel set of servers (on the same box, 24 cores 192GB ram, listening to different ports) operating on identical data sets, and the Dec2016 version is slower than the Jul2015 version (~15-20%). My queries are basic joins between two tables with grouping. I have top running on the same box, and I can see that the Jul2015 server uses all 24 cores, but the Dec2016 server barely cracks 2-3. The MAL plans obtained by explain are totally different (which partly explains the cores being utilized because I can see blocks of 24 commands) and the length of the Dec2016 one is a fraction of the Jul2015 version. I guess the older version was inefficiently using all 24 cores, but the new version efficiently uses a smaller set of cores?? Thanks, Dave

2 2

script-friendly prepare statement
by Roberto Cornacchia 13 Jul '17

13 Jul '17

Hi there, I'm trying to script the execution of (generated) prepared queries. It does not look possible, because the identifier to use for the EXEC statement is returned dynamically by the PREPARE statement: sql>prepare select 1; execute prepared statement using: EXEC 73(...) This can obviously not be scripted. If I tried to parse the output of the prepare statement to retrieve the identifier, this would split PREPARE and EXEC in two different mclient sessions, losing the prepared statement. As far as I know, there is at the moment no syntax available for providing the identifier, something like: sql>prepare myid as select 1; (that wouldn't be a bad idea, would it?) Besides syntax extensions, I was wondering if there is any trick I could use at runtime with the current syntax to retrieve that identifier. Like, is it stored in some system table? Is it available as any undocumented variable? I could not find anything like that, but I might have missed it. Cheers, Roberto

3 4

adding an "insert into" completely changes output from explain
by Anderson, David B 13 Jul '17

13 Jul '17

All, Is it expected behavior if adding "insert into xxxx" at the beginning of a sql query completely alters the output of explain? Without the "insert into", explain returns 1490 tuples, but when I add it, explain only returns 88. I was expecting that the number would go up, i.e. that the original plan would be wrapped in some additional commands which would implement the "insert into" part. Here is my sql: explain insert into aggr select 'CHANNEL', s.product, s.coupon, s.productionyear, s.tpoflag, sum(1), sum(d.scheduled)/1000000, sum(d.realized )/1000000, round(100*(1-power(sum(d.realized)/sum(d.scheduled),12)),3) as cpr from stt s, dyn d where s.loanseqnum = d.loanseqnum group by s.product, s.coupon, s.productionyear, s.tpoflag having sum(d.scheduled) > 0 ;

2 1

Extremely slow order by queries
by Manish gupta 13 Jul '17

13 Jul '17

Hi, I have a scenario where data size is ~3M rows. My query is select C1, C2, C3 from T1 group by C1,C2,C3 This returns result in 30 Sec ( number of groups around 1 M ) While, if I modify the query to select C1, C2, C3 from T1 group by C1,C2,C3 Order by C1, C2 on the same dataset, it takes ~ 1 hour to get result. The platforms I tried is both widows 64 bit and Linux 64 bit, with latest monet installed. Both the machines have 4 cores. And the behavior is same on both the machines. I am really surprised to see overhead of second query. One more observation is that this query occupies only single core, while monet is free to utilize 4 cores. Regards, Manish

2 1

Mserver returns with "memory exhausted in" error if query size exceeds 1.2MB
by Sharma, Sreejith 08 Jul '17

08 Jul '17

Hi, I have encountered scenario where SQL generated exceeds 3mb and it's failing in mserver with an error "memory exhausted in select .......". On further checks I have noticed that the SQL will go through only if SQL size is less than 1.2MB. Has anyone encountered this before? Or is anyone aware of any SQL size limitation in MonetDB? Please help. Regards, Sreejith

1 0

access MapiHdl by using qtag value from C API
by Anton Kravchenko 26 Jun '17

26 Jun '17

Is there a way to access MapiHdl of the query by using its qtag value?

1 0

Re: MonetDB taking lot of storage space
by Stefan Manegold 17 Jun '17

17 Jun '17

Typo in formula:does * columns * (average) width per columnmust readrows * columns * (average) width per column, -------- Original message --------From: Stefan Manegold <stefan.manegold(a)cwi.nl> Date: 6/17/17 12:33 (GMT+01:00) To: Communication channel for MonetDB users <users-list(a)monetdb.org> Subject: Re: MonetDB taking lot of storage space Hi Amir, May I assume "cr" stands for "crore" (c.f.., https://en.m.wikipedia.org/wiki/Crore) I.e. 1 cr = 10 million = 10^7 = 10,000,000 ? The plain expected size of the database assuming only fixed-width column types and ignoring automatically generated indexes, etc.Can be assumed as does * columns * (average) width per column, e.g.,Assuming all 4 byte numerical columns:(4+5) * 10^7 * 10 * 4 byte = 36 * 10^8 ~= 3.6 GBor twice as much for all 8 byte columns. Your 196 GB is far from that. Are you sure there is no other left-over data in your database?As Jennie suggested, did you do a lot of update (insert, delete) operations on these tables?Which (top-level) directories in your dbfarm are the largest? Best,Stefan -------- Original message --------From: Ying Zhang <Y.Zhang(a)cwi.nl> Date: 6/17/17 10:35 (GMT+01:00) To: Communication channel for MonetDB users <users-list(a)monetdb.org> Subject: Re: MonetDB taking lot of storage space Hai, MonetDB doesn’t provide any user definable compressions. What do you mean with “5cr rows”? Compressed Rows? 198GB is indeed pretty big, so it’s expected to contain a lot of table data. If you don’t have variable length columns, it’s easy to compute the expected storage space: #rows * #columns * (sum of the length of each column). Are your database fresh or a lot of insert and delete have been conducted on them? Then, the deletes might have leaved garbages. Did you happen to have put larges BLOBs or CLOBs in some columns? Which MonetDB version and OS do you use? Regards, Jennie > On 17 Jun 2017, at 09:38, Amit Ambekar <amit(a)g-square.in> wrote: > > Hi, > > i have 2 databases in monetdb. first one has 5cr rows and second has 4 cr rows with average 10 columns per table. It takes 196 GB storage space. is it normal ?? most of data consists of numeric values. And How i can compress database size in MonetDB ?? > > _______________________________________________ > users-list mailing list > users-list(a)monetdb.org > https://www.monetdb.org/mailman/listinfo/users-list _______________________________________________ users-list mailing list users-list(a)monetdb.org https://www.monetdb.org/mailman/listinfo/users-list

1 0

Re: MonetDB taking lot of storage space
by Stefan Manegold 17 Jun '17

17 Jun '17

Hi Amir, May I assume "cr" stands for "crore" (c.f.., https://en.m.wikipedia.org/wiki/Crore) I.e. 1 cr = 10 million = 10^7 = 10,000,000 ? The plain expected size of the database assuming only fixed-width column types and ignoring automatically generated indexes, etc.Can be assumed as does * columns * (average) width per column, e.g.,Assuming all 4 byte numerical columns:(4+5) * 10^7 * 10 * 4 byte = 36 * 10^8 ~= 3.6 GBor twice as much for all 8 byte columns. Your 196 GB is far from that. Are you sure there is no other left-over data in your database?As Jennie suggested, did you do a lot of update (insert, delete) operations on these tables?Which (top-level) directories in your dbfarm are the largest? Best,Stefan -------- Original message --------From: Ying Zhang <Y.Zhang(a)cwi.nl> Date: 6/17/17 10:35 (GMT+01:00) To: Communication channel for MonetDB users <users-list(a)monetdb.org> Subject: Re: MonetDB taking lot of storage space Hai, MonetDB doesn’t provide any user definable compressions. What do you mean with “5cr rows”? Compressed Rows? 198GB is indeed pretty big, so it’s expected to contain a lot of table data. If you don’t have variable length columns, it’s easy to compute the expected storage space: #rows * #columns * (sum of the length of each column). Are your database fresh or a lot of insert and delete have been conducted on them? Then, the deletes might have leaved garbages. Did you happen to have put larges BLOBs or CLOBs in some columns? Which MonetDB version and OS do you use? Regards, Jennie > On 17 Jun 2017, at 09:38, Amit Ambekar <amit(a)g-square.in> wrote: > > Hi, > > i have 2 databases in monetdb. first one has 5cr rows and second has 4 cr rows with average 10 columns per table. It takes 196 GB storage space. is it normal ?? most of data consists of numeric values. And How i can compress database size in MonetDB ?? > > _______________________________________________ > users-list mailing list > users-list(a)monetdb.org > https://www.monetdb.org/mailman/listinfo/users-list _______________________________________________ users-list mailing list users-list(a)monetdb.org https://www.monetdb.org/mailman/listinfo/users-list

1 0

MonetDB taking lot of storage space
by Amit Ambekar 17 Jun '17

17 Jun '17

Hi, i have 2 databases in monetdb. first one has 5cr rows and second has 4 cr rows with average 10 columns per table. It takes 196 GB storage space. is it normal ?? most of data consists of numeric values. And How i can compress database size in MonetDB ??

2 1