Loading XML Data into Relational Table

Question

Hello,
I receive an XML file generated from another tool (on Windows), I am trying to create a Linux shell script that will gather the needed XML file from my Linux database server, then have Oracle use this file to load the XML data into a relational table. This activity & data will be needed on an ongoing basis.
I have tried this two ways. First, I loaded the XML document into the database and tried to extract the data directly from the document, but that is not working. Now I want to try to read the data directly from the file on the server via select, however I am not getting any data returned. In the Select statement below, I am simply trying to query the data to see what is returned for my testing.
Create Table ci_results_table (transactionID Varchar2(100), //transactionID should be PrimaryKey but was getting NULL value errors during insert test, so removed PK
message Varchar2(200),
ci Varchar2(50),
processeddate xmltype,
status Varchar2(50),
sourcefile VarChar2(100));
select x.*
from XMLTable(
     'TSPLoadResults/Results'
     PASSING xmltype(bfilename('CMDB_DEVADHOCRESULTS_DIR','LoadResults-HP_146.results.xml'), nls_charset_id('AL32UTF8'))
     COLUMNS
       transactionID Varchar2(100) PATH 'TransactionID',
       Result XMLType PATH 'Result',
       Message Varchar2(200) PATH 'Message',
       PrimaryKey Varchar2(50) PATH 'PrimaryKey',
       ProcessedDate date PATH 'ProcessedDate',
       Status Varchar2(50) PATH 'Status',
       SourceFile VarChar2(100) PATH 'SourceFileName'
   ) x
Eventually I will need to build on this to limit the data returned to those records where SourceFileName is like 'HPDS%' and insert what is returned in to the ci_results_table. Attached is a sample Results XML file I am trying to load, It is named "ResultsTransformedtoUnix" because I used dos2Unix to convert it to Unix which may be right or wrong. (The output file I send out has to be transformed to DOS format before the other application can read it). Original file (before Unix conversion) named in script is also attached.
Please help. Thank you!

Answer 1

Hi,
I see some wrong things in your query.
1) The obvious one, explaining why you're not getting any data : there's a typo in the XQuery expression, it's "Result" not "Results"
2) ProcessedDate cannot be extracted as a date (at least not directly) since it actually represents a timestamp, use TIMESTAMP WITH TIME ZONE datatype and cast back to DATE in the SELECT clause
3) transactionID is an attribute, it must be accessed with '@' (or 'attribute::' axis)
4) If the file encoding is truly ISO-8859-1 as the prolog suggests, then do not use AL32UTF8 but the corresponding charset name : WE8ISO8859P1
Here's the working query :
select x.transactionID
     , x.Message
     , x.Primarykey
     , cast(x.ProcessedDate as date) ProcessDate
     , x.Status
     , x.SourceFile
from XMLTable(
       '/TSPLoadResults/Result'
       PASSING xmltype(bfilename('XML_DIR','LoadResults-HP_146.results.xml'), nls_charset_id('WE8ISO8859P1'))
       COLUMNS
         transactionID Varchar2(100)            PATH '@transactionID',
         Message       Varchar2(200)            PATH 'Message',
         PrimaryKey    Varchar2(50)             PATH 'PrimaryKey',
         ProcessedDate timestamp with time zone PATH 'ProcessedDate',
         Status        Varchar2(50)             PATH 'Status',
         SourceFile    VarChar2(100)            PATH 'SourceFileName'
     ) x
Using this query directly over the file will only perform decently (for large files) on 11.2.0.4 and onwards.
On prior versions, first load the file in a (temporary) XMLType column with Binary XML storage and SELECT from there.
because I used dos2Unix to convert it to Unix which may be right or wrong.
This conversion shouldn't be necessary.