2 Replies Latest reply on Nov 21, 2007 3:20 PM by 609276

    Bulk loading data from several n-triple files

    609276
    609276 Nov 19, 2007 2:06 PM
      Hello,

      What is the best way for bulk-loading data from several n-triple files?
      LUBM50 for example consists of 1,000 files.
      I would highly appreciate if exemplary ctl/sql could be posted!

      Thanks in advance,

      Wolfgang
      • 2182 Views
      • Tags:
        • 1. Re: Bulk loading data from several n-triple files
          Mannamal-Oracle
          Mannamal-Oracle Nov 19, 2007 9:46 PM (in response to 609276)
          Hi Wolfgang,

          Are you looking to load data into 10g or 11g?

          If you are looking at 11g, sample files are located at $ORACLE_HOME/md/demo/network/rdf_demos/ (as specified in the documentation). This location contains a ctl file and a sample sql file that uses that ctl file. You could modify the sample sql file (bulkload.sql) to load multiple files (please refer to the SQL*Loader documentation for more details).

          Detailed instructions on the load process are in section 1.7 of the 11g Oracle Database Semantic Technologies documentation. The Oracle Technology Network includes links to the 11g documentation.

          Melli
          • 2. Re: Bulk loading data from several n-triple files
            609276
            609276 Nov 21, 2007 3:20 PM (in response to Mannamal-Oracle)
            Hi Melli,

            Thanks for your answer. I am loading data into 11g.
            I found the samples. I was also looking for an easy way for uploading all files within a directory, which I found out is not possible directly (i.e. it is not possible to specify a directory in the ctl file). So there are two possible solutions to this:
            1.) Create a list of all files in the directory (on Windows with "dir/b *.nt > files.dat"), prefix every line with "INFILE", surround the filenames with a paragraph and include all that in the ctl file.
            2.) It should also be possible to let a SQL-script create the list ("host dir/b *.nt > files.dat"), load this list into a temp-table and iteratively invoke the sqlldr.

            For me solution 1 worked perfectly.

            However when trying some sample queries (the LUBM test queries) I observed some very strange behaviour of SEM_MATCH (see thread "Missing triples in SEM_MATCH query")

            Wolfgang