(R)mpi and '<U+FEFF>'

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Dear friends,

I got kind of weird problem and no idea how to approach it. First,
what I do.

Again, on a HPC with Rmpi we run a script that does, at very first,
fetches some data from a PostGreSQL DB. The SQL query is read from a
file via 'readLines()'.

What does work.

1) the query itself from psql or pgadmin.
2) the query within the R -script on my local machine and on the cluster

but (and that is the part why I am writing to the list), as soon as
the number of requested cores is that large, that batch system (lsf in
this case) spread the tasks over 2 host, the script fails.

I tried with pure Rmpi, doMPI and foreach and RODBCext and RPostGreSQL
and always encountered the same problem.

Comming back to my subject stated above, the error is:

ERROR:  syntax error at or near "<U+FEFF>WITH"

where my actual query starts with 'WITH'.

Those guys ...

http://postgresql.1045698.n5.nabble.com/BUG-6271-psql-f-reporting-unexpected-syntax-errors-on-first-command-td4940320.html

.... talk about a BOM at the beginning of the line, but somehow this
BOM must be introduced to my file by MPI or Rmpi or whatever .. I have
no idea.

So, any hint is very welcome.

Stefan


- --
Stefan L?dtke

Section 5.4-  Hydrology
Tel.: +49 331 288 2821 <https://rz-vm54.gfz-potsdam.de:443>
Fax: +49 331 288 1570 <https://rz-vm54.gfz-potsdam.de:443>
Email: sluedtke at gfz-potsdam.de

Helmholtz-Zentrum Potsdam
Deutsches GeoForschungsZentrum GFZ
(GFZ German Research Centre for Geoscience)
Stiftung des ?ff. Rechts Land Brandenburg
Telegrafenberg, 14473 Potsdam
- -------------------

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.11 (GNU/Linux)

iQEcBAEBAgAGBQJUPssnAAoJEB5GAbKcg+D8LEIIAL3hxnOdjMVAIBYldP6bcOqk
A2/CJ40luKkx3wVOxBEYwXgKnt+7PEWGS3q6/gEdalvH0nskpm8cThGxhjnfPz6w
sdQKPQzq3SpPv6RTOchohXWRDKcyPWIZOAenJu05QXvnYenmj/NIWCAQKUb7ad+8
0N4o6a/ITAc8x2nfuc6H+EJmxSn79YW/PJRNgmemOedray3FT9HtTx+LcFmsTBlD
H0DkSL+0YJsbssRg7I56JOy3HzM9k3zZN8w/vXFejW56A0aLREB4245SlaAulKrk
sSY3Alz1sKr0GtDFjJBdEacsbzZCpoYuEOSpOv6y/clw6Q5JQxy+Q7c2l17p6sY=
=FDkr
-----END PGP SIGNATURE-----

Hi, Stefan.

I can't replicate the BOM showing up that way even when I'm using a
file which I'm deliberately inserting a BOM into, so this is all
speculation...

But ?wouldn't? ? just strip ?ping? leading "<U+FEFF>" from the ? ?
first element of the vector do the trick? This doesn't resolve the 
reason for the mysterious BOM; that might require more info about
platforms and versions.

-Jonathan

On Wed, Oct 15, 2014 at 3:29 PM, Stefan L?dtke
<sluedtke at gfz-potsdam.de> wrote:

Dear friends,

I got kind of weird problem and no idea how to approach it. First, 
what I do.

Again, on a HPC with Rmpi we run a script that does, at very
first, fetches some data from a PostGreSQL DB. The SQL query is
read from a file via 'readLines()'.

What does work.

1) the query itself from psql or pgadmin. 2) the query within the R
-script on my local machine and on the cluster

but (and that is the part why I am writing to the list), as soon
as the number of requested cores is that large, that batch system
(lsf in this case) spread the tasks over 2 host, the script fails.

I tried with pure Rmpi, doMPI and foreach and RODBCext and
RPostGreSQL and always encountered the same problem.

Comming back to my subject stated above, the error is:

ERROR:  syntax error at or near "<U+FEFF>WITH"

where my actual query starts with 'WITH'.

Those guys ...

http://postgresql.1045698.n5.nabble.com/BUG-6271-psql-f-reporting-unexpected-syntax-errors-on-first-command-td4940320.html

 .... talk about a BOM at the beginning of the line, but somehow
this BOM must be introduced to my file by MPI or Rmpi or whatever
.. I have no idea.

So, any hint is very welcome.

Stefan