| Author |
Message
|
| murdeep |
 Posted: Wed Nov 03, 2004 4:38 pm Post subject: Message Model |
 |
|
Master
Joined: 03 Nov 2004
Posts: 211
|
My company is new to WBI. We are doing proof of concept to determine if it will be used extensively. My manager has assigned me a task that he thinks is easy but I have doubts.
We have a sequential file that is being read by a MQseries application. The entire file is sent as a single message to a queue that a WBI V5 message flow will process. My task is to take the message in WBI and parse it.
I have read the literature and also reviewed support pacs (IA96) but am not being succesfull at my task.
The sequential file is composed of 5 different types of records that appear in any order, any number of times. Each record in the file is a different fixed length. Each record has a starting tag that is unique for each record type. The tag is not the same length for each record. For example 1 record type has a tag 5 bytes long and another record is 3.
I haved tried to create a TDS message format but have run into problems. Here is what I have done.
For each record I created a complex type defined as fixed length. To each of these complex type I defined local elements. I didn't create a local element for the tag. Each of these complex types represents a type of record in the sequential file.
I then create a complex type that will represent the whole message. It is defined as tagged fixed length and to it I add 5 local elements. For each local element I add I set its type to one of the fixed length types I created earlier. Now here is my problem. I want to set the tag for each element added to my complex type but the tags must be all the same length. So i have problems. How do I do this. Isn't tagged fixed length mean a tag followed by a fixed length record?
Regards,
Murdeep |
|
| Back to top |
|
 |
| kman |
| Posted: Wed Nov 03, 2004 11:09 pm Post subject: |
|
|
Partisan
Joined: 21 Jan 2003
Posts: 309
Location: Kuala Lumpur, Malaysia
|
hmmm. I am not sure I understand this. But I would think I would set it up as XML. And since the input file, is actually a message with tags, that should qualifies as XML.
Again, I am not sure of this.
If the tag's name is fixed, then you can parse it. |
|
| Back to top |
|
|
| shanson |
| Posted: Thu Nov 04, 2004 2:56 am Post subject: |
|
|
Partisan
Joined: 17 Oct 2003
Posts: 344
Location: IBM Hursley
|
| TDS Tagged Fixed Length is absolutely the correct approach to start with. The one problem to crack is the variable length tags. Are your tags followed by a 'tag/data separator'? If so you can specify this separator instead of the length. Otherwise I think you will have to model the tag as a local element, and use TDS Use Data Pattern to parse it using regular expressions. |
|
| Back to top |
|
|
| murdeep |
| Posted: Thu Nov 04, 2004 3:56 am Post subject: |
|
|
Master
Joined: 03 Nov 2004
Posts: 211
|
The tags have nothing that indicates there ending point. The Use Data Pattern is an option I did not think about. I will try this option to see if it allows me to recognize the different tags. I don't understand why there isn't a model that allows tagged records where the tags are diferent lengths based on the record. I was looking at group indicators, is that a possible solution?
Regards,
Murdeep |
|
| Back to top |
|
|
| shanson |
| Posted: Thu Nov 04, 2004 4:21 am Post subject: |
|
|
Partisan
Joined: 17 Oct 2003
Posts: 344
Location: IBM Hursley
|
| It's a CATCH 22. The parser needs to identify the records in the bitstream, which it does by looking at the tags. To correctly parse a tag it needs to know when the tag ends - you need a length or a terminating character. |
|
| Back to top |
|
|
| murdeep |
| Posted: Thu Nov 04, 2004 4:34 am Post subject: |
|
|
Master
Joined: 03 Nov 2004
Posts: 211
|
If I tell the parser then tag is ABC the why would I also need to specify a length or a terminator? It seems rather daft that I can't explicily define the tag. This is why I was looking at groups and perhaps a group indicator.
Regards,
Murdeep |
|
| Back to top |
|
|
| shanson |
| Posted: Thu Nov 04, 2004 4:57 am Post subject: |
|
|
Partisan
Joined: 17 Oct 2003
Posts: 344
Location: IBM Hursley
|
| If you follow your logic, how can the parser distinguish a tag of ABC from a tag of AB followed by data C? It can't. The only safe way to parse this is to supply a tag length or a tag data separator, or to use regular expressions where you can look ahead into the data following the tag in order to resolve potential ambiguities. |
|
| Back to top |
|
|
| murdeep |
| Posted: Thu Nov 04, 2004 12:57 pm Post subject: |
|
|
Master
Joined: 03 Nov 2004
Posts: 211
|
I have successfully managed to parse the message using a combination of fixed length and use data pattern. Each record was modelled as a fixed length. I then created a complex type representing the message and modelled it a 'use data pattern'. Each record was added as a local element to the complex type and assigned a 'use data pattern'.
For example one of my records was 150 bytes long. The tag was comprised of the first two fields in the record. These two fileds are 8 bytes in total. Together they would contain 0150REC1. So the 'use data pattern' was denoted as '(0150REC1).{142}' The (0150REC1) tells the parser to look for the occurence of '0150REC1' in the data stream. The .{142} tells the parser to treat the next 142 bytes as the rest of the record. I had to do this kind of patterning for each record type. Once I did this I could parse the entire message.
I still have doubts about how I should be setting the min and max occurences of each record type since I do not know in advance how many of each record type I will get in a message. I also don't know in which order the records will occur. I foresee many hours of testing.
Thanks for everyones help.
Regards,
Murdeep |
|
| Back to top |
|
|
| shanson |
| Posted: Fri Nov 05, 2004 1:23 am Post subject: |
|
|
Partisan
Joined: 17 Oct 2003
Posts: 344
Location: IBM Hursley
|
That's good news Murdeep. As for the occurrences, you have two possibilities:
1) If all occurrences of a given record occur contiguously, then you can simply make the complex type an 'unorderedSet' instead of a 'sequence', and give each of your record elements minOccurs=0 and maxOccurs=-1 (unbounded).
2) If the occurrences of a given record can be interspersed between other types of records, then what you actually have is a repeating choice. Make the complex type a 'choice' then set minOccurs=0 and maxOccurs=-1 on the complex type itself (not the record elements). |
|
| Back to top |
|
|
|
|
