On Sun, 20 Sep 2009 19:54:07 +0100, Roy Brown 
<[log in to unmask]> wrote:

>In message <[log in to unmask]>, Bruce Hobbs
><[log in to unmask]> writing at 13:13:48 in his/her local time
>opines:-
>>On Sun, 20 Sep 2009 00:59:21 -0500, Michael <[log in to unmask]> wrote:
>>
>>>Before you call DBGET with mode=4, you should of previously called dbget
>>>to read a record from this data set 'xxxx',  using mode 1, 2, 3, 5, 6,
>>>or possibly 7. I don't know of any other way to retrieve a valid
>>>RECORD-NUMBER.
>>>
>>I don't recall seeing such a requirement documented. Do you happen to
>>know the page number and version of the manual where you found that
>>requirement?
>
>It's not a technical requirement as such (as I suspect you very well
>know :-) ).
>
It certainly hasn't been my experience nor could I find anything to that effect in the 
manual myself. ;-)

>It's more an indication that except, possibly, in the case of some
>extremely specifically arranged and maintained Image dataset, accessing
>record numbers 'blind' is unlikely to lead to any profitable result.
>
The only profitable result I ever found from such a "blind" access was reading record 
number one in order to initialize a dataset list at the beginning of a program. Of course I 
had to be prepared to ignore a "Directed end of file" if the dataset happened to be empty.

>However, you certainly don't need to have previously read the actual
>record in question; a multi-level bill of materials, say, will store the
>next record at the current level (status elements 9 and 10) while
>exploding the current record (and possibly so on downwards recursively),
>repositioning itself at each returned level with a Mode 4 DBGET on the
>next in chain, and so on.
>
The way I read Michael's posting was not that the record in question, but rather that 
SOME record from that dataset had to have been previously read. I guess I could've 
brought that out better in my original response. :-(

Besides the list initialization I noted and the "specifically arranged and maintained" and 
BOM use cases you mentioned, the only other situation where I can recall using a directed 
read was to re-establish record currency after interrupting the processing of a chain to 
access some other record in the dataset. Of course that's actually very similar to your 
BOM use case.

* To join/leave the list, search archives, change list settings, *
* etc., please visit http://raven.utc.edu/archives/hp3000-l.html *