Re: What I'd Like to See in Delphi

Yannis wrote:
"m. Th." <a@xxxxx> wrote:
Yes, this is the main idea. And this, of course, must be done transparently by the framework even if you wrote *I* want to be able to fetch... etc. IOW for ex. a myClonedDataSet.Locate('Field1', myValue, []) would bring transparently only the records which aren't already cached on the client in the main dataset's buffer.


I do not know if we are talking for the same thing here. As I see it I need to be able to

1) Create multiple cursors/views on a single dataset so different controls will show different portions of the same data eg. using filters, or simple be at different records. This cursors can be read only if this makes things simpler but this views will always show the same data for the same parameters.


Yes.

2) With any dataset connected to a DB I want to be able to re query under conditions. eg MyCloneDataset.Requery(rq_CurrentRecord);
MyCloneDataset.Requery(rq_NewRecordsOnly);
MyCloneDataset.Requery(rq_All);
MyCloneDataset.AppendWith(anSQL);


Ok, if you want - I'm not against these, but I see it like a component like, let's say a TDBGrid. When your main TDataSet descendant is updated for whatever reason, then your Clone is updated (using a messaging system for ex.) from the TDataSet's _buffer_. In fact this is the main value of the component: to minimize the roundtrips to the server (or to another n-tier, for example).

Or IOW, The TCloneDataSet is expected to work in the same way in which TDBGrid issues a fetch request to TDataSet and is TDataSet's responsibility to decide from where it fetches its records (local buffer/cache or a n-tiered DB backend). This of course, imply that we rely on TDataSet descendant which is supposed to know how to manage its internal bi-directional cache in order to give in the shortest time the records requested by the clones.

I do not expect the Dataset to have any internal sync mechanism. In fact I just ask for the opposite, instead of writing a sync mechanism which probably will not fit most of the programs give me the ability to create one by extending the existing controls to allow for it. I do not like each locate/search/filter start hitting the Server /remote connection to check if there are new or updated records, I might be inside a tight loop at the moment calculating something for the existing data only.


See above.

regards
Yannis


HTH,

--

m. th.
.

Relevant Pages

  • Re: Large Table Deletes
    ... selects 8192 rowids (all that fit into the until recently maximum fetch ... buffer of 32K - that has been expanded to 4GB fairly recently and I'm in the ...
    (comp.databases.informix)
  • Re: Non blocking spin lock..
    ... then later complete the fetch of buffer_to_read. ... from buffer'. ... fetch data across a pointer dereference. ... Write ordering is *not* part of the x86 specification. ...
    (comp.programming.threads)
  • Re: An analogic ALU
    ... exemple) an then fetch the result using a DAC followed by a buffer. ...
    (comp.arch.arithmetic)
  • Re: recover from failed make build
    ... Dumb question - is current defined as pwd? ... iow if I cd ... /usr/ports/www/Apache2 and then make distclean;make fetch will that attempt ...
    (freebsd-questions)
  • Re: WHY WOULD ANYBODY TRUST A HIPPY......
    ... I doubt you'll get a reply from fetch on that one. ... being built such that he can no longer enjoy that buffer and privacy. ... He fails to grasp the notion, though, that the neighborhod that he ... lives in now was, at one time, new and undoubtedly encroached on ...
    (rec.sport.disc)