Best strategy for MU data entry form with multiple record subform

Hi All

Just how do developers sensibly handle record locking conflicts in this
situation:

I have a multiuser FE/BE A2002 database. There is a single form for both
data entry and subsequent re-editing. The main form is bound to the Sales
Order header table, and the child form to the Line detail table. I decided to
use a transaction so that the user may enter/edit the entire order and
save/abandon it entirely (which seems to be 'universal' behaviour for this
type of interface). This is of course where it comes unstuck! I assumed
(foolishly) that I could use optimistic, record-level locking and get away
with it. The whole thing works beautifully for a single user, but
practically, only one user at a time may enter/edit data because the
possibilty of multiple dirty rows in the subform appears to force Jet to
switch to page-level locking...help!!
I'm using DAO bound recordsets, because I'm only really familiar with DAO -
could ADO help me out here?
I wondered about using an unbound form, or bound form with unbound controls,
but I can't imagine how I could display the multiple detail rows without the
bound subform in datasheet mode.

This must be a classic programming challenge - forgive my ignorance!

Any and all suggestions appreciated.

Ali Kwok

Why use a transaction? That seems to be the place where you are making this
more difficult that it needs to be.

Instead, set up the relation with a cascading delete between the main order
table and the line items table. If the user "abandons" the order entry (i.e.
deletes the main form entry), the related line order items are deleted as
well. So, no transaction is needed for deletes.

As the user adds new records to the line items, no transaction is needed.

If the user edits records in the line items, no transaction is needed.

You're done, and you're avoided the whole puzzle of trying to manage
simultaneous uncommitted transactions beyond the single-record buffering and
optimistic writes that Access does so well.

--
Allen Browne - Microsoft MVP. Perth, Western Australia.
Tips for Access users - http://allenbrowne.com/tips.html
Reply to group, rather than allenbrowne at mvps dot org.

"AliKwok" wrote in message
...

Just how do developers sensibly handle record locking conflicts in this
situation:

I have a multiuser FE/BE A2002 database. There is a single form for both
data entry and subsequent re-editing. The main form is bound to the Sales
Order header table, and the child form to the Line detail table. I decided
to
use a transaction so that the user may enter/edit the entire order and
save/abandon it entirely (which seems to be 'universal' behaviour for this
type of interface). This is of course where it comes unstuck! I assumed
(foolishly) that I could use optimistic, record-level locking and get away
with it. The whole thing works beautifully for a single user, but
practically, only one user at a time may enter/edit data because the
possibilty of multiple dirty rows in the subform appears to force Jet to
switch to page-level locking...help!!
I'm using DAO bound recordsets, because I'm only really familiar with
DAO -
could ADO help me out here?
I wondered about using an unbound form, or bound form with unbound
controls,
but I can't imagine how I could display the multiple detail rows without
the
bound subform in datasheet mode.

This must be a classic programming challenge - forgive my ignorance!

Any and all suggestions appreciated.

Ali Kwok

Thanks for your help, Allen - it's not the first time.

I follow the cascading deletes logic OK, but I also need users to be able to
abandon ANY edits - whether or not they have moved between rows in the
subform. I don't see how I can avoid committing such changes without using a
transaction(?)

The 'complete reversion' functionality is needed to enable rework of order
quantities etc to test the effects on pricing and delivery time - while the
customer waits on the phone - with the confidence to be able to leave the
original order untouched if necessary.

Thank you

Ali

"Allen Browne" wrote:

Why use a transaction? That seems to be the place where you are making this
more difficult that it needs to be.

Instead, set up the relation with a cascading delete between the main order
table and the line items table. If the user "abandons" the order entry (i.e.
deletes the main form entry), the related line order items are deleted as
well. So, no transaction is needed for deletes.

As the user adds new records to the line items, no transaction is needed.

If the user edits records in the line items, no transaction is needed.

You're done, and you're avoided the whole puzzle of trying to manage
simultaneous uncommitted transactions beyond the single-record buffering and
optimistic writes that Access does so well.

--
Allen Browne - Microsoft MVP. Perth, Western Australia.
Tips for Access users - http://allenbrowne.com/tips.html
Reply to group, rather than allenbrowne at mvps dot org.

"AliKwok" wrote in message
...

Just how do developers sensibly handle record locking conflicts in this
situation:

I have a multiuser FE/BE A2002 database. There is a single form for both
data entry and subsequent re-editing. The main form is bound to the Sales
Order header table, and the child form to the Line detail table. I decided
to
use a transaction so that the user may enter/edit the entire order and
save/abandon it entirely (which seems to be 'universal' behaviour for this
type of interface). This is of course where it comes unstuck! I assumed
(foolishly) that I could use optimistic, record-level locking and get away
with it. The whole thing works beautifully for a single user, but
practically, only one user at a time may enter/edit data because the
possibilty of multiple dirty rows in the subform appears to force Jet to
switch to page-level locking...help!!
I'm using DAO bound recordsets, because I'm only really familiar with
DAO -
could ADO help me out here?
I wondered about using an unbound form, or bound form with unbound
controls,
but I can't imagine how I could display the multiple detail rows without
the
bound subform in datasheet mode.

This must be a classic programming challenge - forgive my ignorance!

Any and all suggestions appreciated.

Ali Kwok

Okay. You have a specific goal in mind.

But there still might be an easier way to handle the stock quantity issues.
One way is to handle the stock quantities dynamically, i.e. don't store the
stock values, beyond a known point such as a stock take or start-of-day
value. Another is illustrated by John Viescas in Building MS Access
Applications (Microsoft Press, 2003), which includes a sample inventory
management application.

--
Allen Browne - Microsoft MVP. Perth, Western Australia.
Tips for Access users - http://allenbrowne.com/tips.html
Reply to group, rather than allenbrowne at mvps dot org.

"AliKwok" wrote in message
...
Thanks for your help, Allen - it's not the first time.

I follow the cascading deletes logic OK, but I also need users to be able
to
abandon ANY edits - whether or not they have moved between rows in the
subform. I don't see how I can avoid committing such changes without using
a
transaction(?)

The 'complete reversion' functionality is needed to enable rework of order
quantities etc to test the effects on pricing and delivery time - while
the
customer waits on the phone - with the confidence to be able to leave the
original order untouched if necessary.

Thank you

Ali

"Allen Browne" wrote:

Why use a transaction? That seems to be the place where you are making
this
more difficult that it needs to be.

Instead, set up the relation with a cascading delete between the main
order
table and the line items table. If the user "abandons" the order entry
(i.e.
deletes the main form entry), the related line order items are deleted as
well. So, no transaction is needed for deletes.

As the user adds new records to the line items, no transaction is needed.

If the user edits records in the line items, no transaction is needed.

You're done, and you're avoided the whole puzzle of trying to manage
simultaneous uncommitted transactions beyond the single-record buffering
and
optimistic writes that Access does so well.

"AliKwok" wrote in message
...

Just how do developers sensibly handle record locking conflicts in this
situation:

I have a multiuser FE/BE A2002 database. There is a single form for
both
data entry and subsequent re-editing. The main form is bound to the
Sales
Order header table, and the child form to the Line detail table. I
decided
to
use a transaction so that the user may enter/edit the entire order and
save/abandon it entirely (which seems to be 'universal' behaviour for
this
type of interface). This is of course where it comes unstuck! I assumed
(foolishly) that I could use optimistic, record-level locking and get
away
with it. The whole thing works beautifully for a single user, but
practically, only one user at a time may enter/edit data because the
possibilty of multiple dirty rows in the subform appears to force Jet
to
switch to page-level locking...help!!
I'm using DAO bound recordsets, because I'm only really familiar with
DAO -
could ADO help me out here?
I wondered about using an unbound form, or bound form with unbound
controls,
but I can't imagine how I could display the multiple detail rows
without
the
bound subform in datasheet mode.

This must be a classic programming challenge - forgive my ignorance!

If the Order is abandonded (deleted), the SCENARIOS are moot. The
records would be deleted. I would probably approach it from the
standpoint of creating a duplicate purchase order which can be changed
around in any such way. If you don't already have a [STATUS] field in
the SALES ORDER header table, add one and create a new [STATUS] such as
'SCENRIO' or 'DUPLICATE' something that separates it out from being a
LIVE Sales Order. If the user wants to discard the changes, you would
just delete the duplicate sales order. If the user wants to save the
changes, you would run an UPDATE query to update the line items for the
LIVE sales order to the values of the line items for the DUPLICATE sales
order. Assuming of course that you want to keep the SALES ORDER NUMBER
the same. Otherwise, it would be a matter of updating the [STATUS] of
the first to 'Cancelled' and updating the status of the duplicate to
'Active/Live'.

AliKwok wrote:
Thanks for your help, Allen - it's not the first time.

I follow the cascading deletes logic OK, but I also need users to be able to
abandon ANY edits - whether or not they have moved between rows in the
subform. I don't see how I can avoid committing such changes without using a
transaction(?)

The 'complete reversion' functionality is needed to enable rework of order
quantities etc to test the effects on pricing and delivery time - while the
customer waits on the phone - with the confidence to be able to leave the
original order untouched if necessary.

Thank you

Ali

"Allen Browne" wrote:


Why use a transaction? That seems to be the place where you are making this
more difficult that it needs to be.

Instead, set up the relation with a cascading delete between the main order
table and the line items table. If the user "abandons" the order entry (i.e.
deletes the main form entry), the related line order items are deleted as
well. So, no transaction is needed for deletes.

As the user adds new records to the line items, no transaction is needed.

If the user edits records in the line items, no transaction is needed.

You're done, and you're avoided the whole puzzle of trying to manage
simultaneous uncommitted transactions beyond the single-record buffering and
optimistic writes that Access does so well.

--
Allen Browne - Microsoft MVP. Perth, Western Australia.
Tips for Access users - http://allenbrowne.com/tips.html
Reply to group, rather than allenbrowne at mvps dot org.

"AliKwok" wrote in message
...

Just how do developers sensibly handle record locking conflicts in this
situation:

I have a multiuser FE/BE A2002 database. There is a single form for both
data entry and subsequent re-editing. The main form is bound to the Sales
Order header table, and the child form to the Line detail table. I decided
to
use a transaction so that the user may enter/edit the entire order and
save/abandon it entirely (which seems to be 'universal' behaviour for this
type of interface). This is of course where it comes unstuck! I assumed
(foolishly) that I could use optimistic, record-level locking and get away
with it. The whole thing works beautifully for a single user, but
practically, only one user at a time may enter/edit data because the
possibilty of multiple dirty rows in the subform appears to force Jet to
switch to page-level locking...help!!
I'm using DAO bound recordsets, because I'm only really familiar with
DAO -
could ADO help me out here?
I wondered about using an unbound form, or bound form with unbound
controls,
but I can't imagine how I could display the multiple detail rows without
the
bound subform in datasheet mode.

This must be a classic programming challenge - forgive my ignorance!

Any and all suggestions appreciated.

Ali Kwok

(Additional Comment on my last post)...

There is no 'BEST' strategy. There are multiple ways of accomplishing
the same task. The strategy choosen should be the one that works in
light of your specific requirements and is consistent with the overall
design & approach of your solution.

AliKwok wrote:
Thanks for your help, Allen - it's not the first time.

I follow the cascading deletes logic OK, but I also need users to be able to
abandon ANY edits - whether or not they have moved between rows in the
subform. I don't see how I can avoid committing such changes without using a
transaction(?)

The 'complete reversion' functionality is needed to enable rework of order
quantities etc to test the effects on pricing and delivery time - while the
customer waits on the phone - with the confidence to be able to leave the
original order untouched if necessary.

Thank you

Ali

"Allen Browne" wrote:


Why use a transaction? That seems to be the place where you are making this
more difficult that it needs to be.

Instead, set up the relation with a cascading delete between the main order
table and the line items table. If the user "abandons" the order entry (i.e.
deletes the main form entry), the related line order items are deleted as
well. So, no transaction is needed for deletes.

As the user adds new records to the line items, no transaction is needed.

If the user edits records in the line items, no transaction is needed.

You're done, and you're avoided the whole puzzle of trying to manage
simultaneous uncommitted transactions beyond the single-record buffering and
optimistic writes that Access does so well.

--
Allen Browne - Microsoft MVP. Perth, Western Australia.
Tips for Access users - http://allenbrowne.com/tips.html
Reply to group, rather than allenbrowne at mvps dot org.

"AliKwok" wrote in message
...

Just how do developers sensibly handle record locking conflicts in this
situation:

I have a multiuser FE/BE A2002 database. There is a single form for both
data entry and subsequent re-editing. The main form is bound to the Sales
Order header table, and the child form to the Line detail table. I decided
to
use a transaction so that the user may enter/edit the entire order and
save/abandon it entirely (which seems to be 'universal' behaviour for this
type of interface). This is of course where it comes unstuck! I assumed
(foolishly) that I could use optimistic, record-level locking and get away
with it. The whole thing works beautifully for a single user, but
practically, only one user at a time may enter/edit data because the
possibilty of multiple dirty rows in the subform appears to force Jet to
switch to page-level locking...help!!
I'm using DAO bound recordsets, because I'm only really familiar with
DAO -
could ADO help me out here?
I wondered about using an unbound form, or bound form with unbound
controls,
but I can't imagine how I could display the multiple detail rows without
the
bound subform in datasheet mode.

This must be a classic programming challenge - forgive my ignorance!

Any and all suggestions appreciated.

Ali Kwok

First, using transaction only works for you recordset code, and DOES NOT
work for forms.

In other words, starting a transaction has not relation to a form.


the possibilty of multiple dirty rows in the subform appears to force Jet
to
switch to page-level locking...help!!

Not to my knowledge at all does the above happen. However, I don't see why
the above would be a problem, or related to this in any way. You don't
mention what version of ms-access, but the last 3 versions of ms-access do
have record locking as opposed to page locking anyway. (so, anything after
a97 would not be a problem. However, even a97 is not a problem, since there
is NO REASON TO HAVE ANY kind of locking enable, or set for the CHILD
TABLES!!).

So, this issue is moot, since why bother locking the child records in ANY
WAY at all? The ONLY way you are EVER going to edit a child record is by
first finding, and bringing the master record/form. The child records will
then display. So, you can't get to the child records unless you edit/find
the MASTER record. (so, really, you only need to lock the master record). I
see ZERO reason as to why you have and child record locking in the first
place.

As mentioned, actually all of the above may be moot, as transactions have
NOTHING to do with forms anyway.


This must be a classic programming challenge - forgive my ignorance!

Allowing a bail out of a master/child forms is something that ms-access does
not do well. You *can* bind recordsets to a form that are wrapped in a
transaction, but this assumes you have a existing design that will allow
this (and that is not normal the case).

So, keep in mind the issue of transactions, the issue of forms, the issue of
locking...as they are all in face separate issues.

Having said the above, you *can* build your own recordsets in code, and then
ASSIGN these reocrdets to the form, and sub-forms reocrdset property. The
link master/child settings do not work when you do this, but it is simple
matter to put the assigning of the parent id in the child form in the before
insert event of the child form

Me!contact_id = Me.Parent!ContactID

So, you can wrap a form in a transaction, but you will thus need to load up
the one record into the master reocrdset, and load up the child records into
a child reocrdset. This likely also means you will need some setup for the
"adding" of records..

--
Albert D. Kallal (Access MVP)
Edmonton, Alberta Canada

http://www.members.shaw.ca/AlbertKallal

Thanks for your reply Albert.

I have indeed wrapped the form in a transaction by programmatically
assigning the recordsets, and as I wrote, it was working fine for a single
user, but appears to lock whole pages when 2 users are attempting to enter
new records, or edit old ones. Version is A2002, as I wrote.
Since my settings are No Locks/Record-level locking and the recordsets'
locking setting is Optimistic, I can only suggest that 1) the Transaction
must override the optimistic setting, and 2) the potential for multiple child
records within the scope of the Transaction must override the record-level
locking. This is only my hypothesis to explain the observed behaviour - I
havn't had time to test it!

Thanks for your time and thought.

Ali

"Albert D.Kallal" wrote:

First, using transaction only works for you recordset code, and DOES NOT
work for forms.

In other words, starting a transaction has not relation to a form.


the possibilty of multiple dirty rows in the subform appears to force Jet
to
switch to page-level locking...help!!

Not to my knowledge at all does the above happen. However, I don't see why
the above would be a problem, or related to this in any way. You don't
mention what version of ms-access, but the last 3 versions of ms-access do
have record locking as opposed to page locking anyway. (so, anything after
a97 would not be a problem. However, even a97 is not a problem, since there
is NO REASON TO HAVE ANY kind of locking enable, or set for the CHILD
TABLES!!).

So, this issue is moot, since why bother locking the child records in ANY
WAY at all? The ONLY way you are EVER going to edit a child record is by
first finding, and bringing the master record/form. The child records will
then display. So, you can't get to the child records unless you edit/find
the MASTER record. (so, really, you only need to lock the master record). I
see ZERO reason as to why you have and child record locking in the first
place.

As mentioned, actually all of the above may be moot, as transactions have
NOTHING to do with forms anyway.


This must be a classic programming challenge - forgive my ignorance!

Allowing a bail out of a master/child forms is something that ms-access does
not do well. You *can* bind recordsets to a form that are wrapped in a
transaction, but this assumes you have a existing design that will allow
this (and that is not normal the case).

So, keep in mind the issue of transactions, the issue of forms, the issue of
locking...as they are all in face separate issues.

Having said the above, you *can* build your own recordsets in code, and then
ASSIGN these reocrdets to the form, and sub-forms reocrdset property. The
link master/child settings do not work when you do this, but it is simple
matter to put the assigning of the parent id in the child form in the before
insert event of the child form

Me!contact_id = Me.Parent!ContactID

So, you can wrap a form in a transaction, but you will thus need to load up
the one record into the master reocrdset, and load up the child records into
a child reocrdset. This likely also means you will need some setup for the
"adding" of records..

--
Albert D. Kallal (Access MVP)
Edmonton, Alberta Canada

http://www.members.shaw.ca/AlbertKallal

"AliKwok" wrote in message
...

I can only suggest that 1) the Transaction
must override the optimistic setting, and 2) the potential for multiple
child
records within the scope of the Transaction must override the record-level
locking. This is only my hypothesis to explain the observed behaviour - I
havn't had time to test it!

I seem to recall that some issues can arise. First, I would remove the forms
(and sub-forms) recordsouce. (I assume you built the forms with a bound
table/query). You then REMOVE the data source, since you are now using
recordsets.

If you done the above, then I perhaps there is some issue with locking. I
have not used the transaction idea in production, but I have test a form,
and it did seem to work ok (but, I did notice a lock on the record that
appeared *after* I did a commit. So, there certainly might be some lock
issue.

I would also make sure you are running a split database if not already.


--
Albert D. Kallal (Access MVP)
Edmonton, Alberta Canada

http://www.members.shaw.ca/AlbertKallal

Have you considered the suggestion that I made in another post? In the
amount of time spent banging your head on the wall to get your fax
software to work, you could have easily driven the fax across the
street. (ie Sometimes is better to consider a different approach)

AliKwok wrote:
Thanks for your reply Albert.

I have indeed wrapped the form in a transaction by programmatically
assigning the recordsets, and as I wrote, it was working fine for a single
user, but appears to lock whole pages when 2 users are attempting to enter
new records, or edit old ones. Version is A2002, as I wrote.
Since my settings are No Locks/Record-level locking and the recordsets'
locking setting is Optimistic, I can only suggest that 1) the Transaction
must override the optimistic setting, and 2) the potential for multiple child
records within the scope of the Transaction must override the record-level
locking. This is only my hypothesis to explain the observed behaviour - I
havn't had time to test it!

Thanks for your time and thought.

Ali

"Albert D.Kallal" wrote:


First, using transaction only works for you recordset code, and DOES NOT
work for forms.

In other words, starting a transaction has not relation to a form.



the possibilty of multiple dirty rows in the subform appears to force Jet
to
switch to page-level locking...help!!

Not to my knowledge at all does the above happen. However, I don't see why
the above would be a problem, or related to this in any way. You don't
mention what version of ms-access, but the last 3 versions of ms-access do
have record locking as opposed to page locking anyway. (so, anything after
a97 would not be a problem. However, even a97 is not a problem, since there
is NO REASON TO HAVE ANY kind of locking enable, or set for the CHILD
TABLES!!).

So, this issue is moot, since why bother locking the child records in ANY
WAY at all? The ONLY way you are EVER going to edit a child record is by
first finding, and bringing the master record/form. The child records will
then display. So, you can't get to the child records unless you edit/find
the MASTER record. (so, really, you only need to lock the master record). I
see ZERO reason as to why you have and child record locking in the first
place.

As mentioned, actually all of the above may be moot, as transactions have
NOTHING to do with forms anyway.



This must be a classic programming challenge - forgive my ignorance!

Allowing a bail out of a master/child forms is something that ms-access does
not do well. You *can* bind recordsets to a form that are wrapped in a
transaction, but this assumes you have a existing design that will allow
this (and that is not normal the case).

So, keep in mind the issue of transactions, the issue of forms, the issue of
locking...as they are all in face separate issues.

Having said the above, you *can* build your own recordsets in code, and then
ASSIGN these reocrdets to the form, and sub-forms reocrdset property. The
link master/child settings do not work when you do this, but it is simple
matter to put the assigning of the parent id in the child form in the before
insert event of the child form

Me!contact_id = Me.Parent!ContactID

So, you can wrap a form in a transaction, but you will thus need to load up
the one record into the master reocrdset, and load up the child records into
a child reocrdset. This likely also means you will need some setup for the
"adding" of records..

--
Albert D. Kallal (Access MVP)
Edmonton, Alberta Canada

http://www.members.shaw.ca/AlbertKallal