[Monetdb-developers] [PATCH] MapiClient bugfix && copy from stdin
I wanted to open a feature request for this, but it seems I forgot my sourceforge password. I attach two patches (cvs -uN): ===== interactive.patch It fixes what I would consider a bug in MapiClient: if the '-s' and '-i' switches are both given, and the argument to '-s' is not finished in the command line (mid->active is not null, i.e, if the -s is a copyfrom construct), MapiClient will segfault instead of finishing the command from stdin. before: $ MapiClient -l sql -i -s "copy 1 records into v from stdin" sql>dfjk Segmentation fault after: $ MapiClient -l sql -i -s "copy 1 records into v from stdin" more>dfjk [ 1 ] sql> There is a remaining issue if the copyfrom ends in a semicolon in the -s switch, because MapiClient will insert another semicolon after that... I did not touch that piece of code buecasue from thhe comments around it, it seemed like it was done on purpose. ======= copyfrom.patch Allows a copy from stdin without specifying the number of records. End of input will be marked with a '\.' on an empty line, as in postgres. Additionally, if stdin ends (or a ctrl-D is received), MapiClient will feed the end-of-input to the backend. ===== Thus, with both patches applied, it should be now possible to do myprocess | MapiClient -l sql -i -s "copy into v from stdin" If interactive is not applied, it should still be possible to do: ( echo "copy into v from stdin;"; myprocess ) | MapiClient -l sql Please tell me how to proceed. Cheers, Zarrabeitia.
On Wed, Aug 09, 2006 at 03:13:13PM -0400, Luis Zarrabeitia wrote:
I wanted to open a feature request for this, but it seems I forgot my sourceforge password.
Many thanks for your patches.
I attach two patches (cvs -uN):
I aplied the interactive patch. Not the '\.' copyfrom patch, as it changes a sql specific issue in MapiClient which is a more general tool (also for xquery and mil/mal). I made some more changes in the sql frontend and ascii_io library to solve the proces | MapiClient problem with out '\.' (just end of file should do). Niels
===== interactive.patch It fixes what I would consider a bug in MapiClient: if the '-s' and '-i' switches are both given, and the argument to '-s' is not finished in the command line (mid->active is not null, i.e, if the -s is a copyfrom construct), MapiClient will segfault instead of finishing the command from stdin.
before: $ MapiClient -l sql -i -s "copy 1 records into v from stdin" sql>dfjk Segmentation fault
after: $ MapiClient -l sql -i -s "copy 1 records into v from stdin" more>dfjk [ 1 ] sql>
There is a remaining issue if the copyfrom ends in a semicolon in the -s switch, because MapiClient will insert another semicolon after that... I did not touch that piece of code buecasue from thhe comments around it, it seemed like it was done on purpose.
======= copyfrom.patch Allows a copy from stdin without specifying the number of records. End of input will be marked with a '\.' on an empty line, as in postgres. Additionally, if stdin ends (or a ctrl-D is received), MapiClient will feed the end-of-input to the backend. =====
Thus, with both patches applied, it should be now possible to do
myprocess | MapiClient -l sql -i -s "copy into v from stdin"
If interactive is not applied, it should still be possible to do:
( echo "copy into v from stdin;"; myprocess ) | MapiClient -l sql
Please tell me how to proceed.
Cheers,
Zarrabeitia.
------------------------------------------------------------------------- Using Tomcat but need to do more? Need to support web services, security? Get stuff done quickly with pre-integrated technology to make your job easier Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642 _______________________________________________ Monetdb-developers mailing list Monetdb-developers@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/monetdb-developers
-- Niels Nes, Centre for Mathematics and Computer Science (CWI) Kruislaan 413, 1098 SJ Amsterdam, The Netherlands room C0.02, phone ++31 20 592-4098, fax ++31 20 592-4312 url: http://www.cwi.nl/~niels e-mail: Niels.Nes@cwi.nl
Many thanks for your patches.
A pleasure.
I attach two patches (cvs -uN):
I aplied the interactive patch.
Cool!
Not the '\.' copyfrom patch, as it changes a sql specific issue in MapiClient which is a more general tool (also for xquery and mil/mal). I made some more changes in the sql frontend and ascii_io library to solve the proces | MapiClient problem with out '\.' (just end of file should do).
I imagined that \. could cause some semantic trouble. I went ahead anyway because I wanted "myprocess |" (or a "cat myfile |") to be able to execute more than one copyfrom and \. seemed a good choice (postgres). I shouldn't have touched the MapiClient, though. I like that you fixed it. Would be acceptable to introduce the \. without touching the MapiClient? (changes could be confined to ascii_io.mx)
Niels
Zarrabeitia.
On 16-08-2006 17:30:30 -0400, Luis Zarrabeitia wrote:
I imagined that \. could cause some semantic trouble. I went ahead anyway because I wanted "myprocess |" (or a "cat myfile |") to be able to execute more than one copyfrom and \. seemed a good choice (postgres). I shouldn't have touched the MapiClient, though. I like that you fixed it. Would be acceptable to introduce the \. without touching the MapiClient? (changes could be confined to ascii_io.mx)
In fact, the PostgreSQL dump format contains other things like \NULL or \0 (not sure which one) together with the \. to "terminate" the dump. Supporting this (in the back-end) probably means writing/patching another ascii_io called postgres_io or something :) I haven't looked into it too deep, but maybe this is for now with the current state one bridge too far.
On Wed, Aug 16, 2006 at 05:30:30PM -0400, Luis Zarrabeitia wrote:
Not the '\.' copyfrom patch, as it changes a sql specific issue in MapiClient which is a more general tool (also for xquery and mil/mal). I made some more changes in the sql frontend and ascii_io library to solve the proces | MapiClient problem with out '\.' (just end of file should do).
I imagined that \. could cause some semantic trouble. I went ahead anyway because I wanted "myprocess |" (or a "cat myfile |") to be able to execute more than one copyfrom and \. seemed a good choice (postgres). I shouldn't have touched the MapiClient, though. I like that you fixed it. Would be acceptable to introduce the \. without touching the MapiClient? (changes could be confined to ascii_io.mx)
Adding the '\.' handling in ascii_io sounds fine with me. As long as its only an extra bonus, ie ascii_io should end with reaching the number of records, end of file or hitting \. (be sure to see if it doesn't give problems with single non quoted string columns. Niels
Niels
Zarrabeitia.
-- Niels Nes, Centre for Mathematics and Computer Science (CWI) Kruislaan 413, 1098 SJ Amsterdam, The Netherlands room C0.02, phone ++31 20 592-4098, fax ++31 20 592-4312 url: http://www.cwi.nl/~niels e-mail: Niels.Nes@cwi.nl
participants (3)
-
Fabian Groffen -
Luis Zarrabeitia -
Niels Nes