Temporarily remove referential integrity checks through VBA?

[Andrew R]

Hi

I'm creating a series of forms, each with with around 15-20 text boxes.
The text boxes will show data from tables, but are unbound to make them
more flexible.

I want the form to be used for both adding new data and modifying
existing data. I have created a save button on the form.

When the user clicks the save button, the code checks to see if there
is a value in the ID text box. If not, it assumes that the record is
new and uses an "Insert into..." SQL statement as the ADODB command
text. This works fine.

However, if there is already an ID value present, the logic is that the
record already exists, and therefore we are doing an update. Creating a
SQL statement along the lines of "UPDATE tblCustomers SET strFirstName
= " & txtFirstName.text & ", strLastName=" & txtLastName.text
...... and so on seems to be a lot of coding to update perhaps 14 fields
which haven't changed in addition to the one that may have.

Therefore, my question is this: Is there an easy way to detect which
field has been changed (without writing code behind each text box,
given that Access doesn't allow control arrays... It probably seems
like I'm being lazy, but with probably 8 or so forms, each with 15-20
text boxes, that would be a lot of work!)

Assuming not, my other thought was that it would be much simpler to
simply execute 2 commands -
"Delete from tblCustomers where ID=" & txtID.text
and then run the insert statement as before. However, this will of
course be prevented by the referential integrity constraints on the
table, thus my question - is there a way through the code to turn off
checks, run the delete and insert statements, then turn them on again?

Apologies for being long-winded! Any help gratefully received!

Regards
Andrew Richards

 

[Arno R]

Hi

I'm creating a series of forms, each with with around 15-20 text boxes.
The text boxes will show data from tables, but are unbound to make them
more flexible.

I want the form to be used for both adding new data and modifying
existing data. I have created a save button on the form.

When the user clicks the save button, the code checks to see if there
is a value in the ID text box. If not, it assumes that the record is
new and uses an "Insert into..." SQL statement as the ADODB command
text. This works fine.

However, if there is already an ID value present, the logic is that the
record already exists, and therefore we are doing an update. Creating a
SQL statement along the lines of "UPDATE tblCustomers SET strFirstName
= " & txtFirstName.text & ", strLastName=" & txtLastName.text
..... and so on seems to be a lot of coding to update perhaps 14 fields
which haven't changed in addition to the one that may have.

Therefore, my question is this: Is there an easy way to detect which
field has been changed (without writing code behind each text box,
given that Access doesn't allow control arrays... It probably seems
like I'm being lazy, but with probably 8 or so forms, each with 15-20
text boxes, that would be a lot of work!)

Assuming not, my other thought was that it would be much simpler to
simply execute 2 commands -
"Delete from tblCustomers where ID=" & txtID.text
and then run the insert statement as before. However, this will of
course be prevented by the referential integrity constraints on the
table, thus my question - is there a way through the code to turn off
checks, run the delete and insert statements, then turn them on again?

Apologies for being long-winded! Any help gratefully received!

Regards
Andrew Richards

Removing RI is *not* the way to go
Deleting the tblCustomers-record is *not* the way to go

These are both very bad and dangerous idea's IMO

What is wrong with bound textboxes??
You could edit, add, delete, cancel changes and so on.
You could save yourself a lot of coding indeed! and... you would not have these problems.

Arno R

 

[Terry Kreft]

You either use a bound form and save yourself a bunch of coding but you are
tied to the bound way of doing things

or

You use unbound and have to do a bunch of coding but you have the
flexibility to do exactly what you want.

You've chosen unbound and now you have to bite the bullet and write the
code.

Try thinking about writing a SQL builder function which you can use from any
of your forms. You pass the tablename and the values from your fields and
let the SQL builder function build the SQL for the update. Do it once, get
it right and then just use it.

 

['69 Camaro]

Hi, Andrew.

The text boxes will show data from tables, but are unbound to make them
more flexible.

That extra flexibility comes at the risk of compromising data integrity and
at a cost of extra work on the database developer's part. Are you sure that
you can't do what you need to do with bound forms?

Is there an easy way to detect which
field has been changed (without writing code behind each text box,

Yes. You can do a visual inpection of the records first. But SQL (or even
VBA code) to update these records would be far more efficient if you ask me.

It probably seems
like I'm being lazy, but with probably 8 or so forms, each with 15-20
text boxes, that would be a lot of work!)

Uh, . . . isn't that what you're getting paid for? To do hours and hours of
work? Every day you come to work? ;-)

is there a way through the code to turn off
checks, run the delete and insert statements, then turn them on again?

Jet doesn't have deferred constraints like client/server databases do, so
you'll have to drop the constraints, not just "turn them off." Honestly, an
update query is the best way to handle this, but if you want to do it the
hard way, you can delete, then insert the appropriate records, but you'll
have to take special precautions.

First, back up the database in case something goes wrong. Next, create a
table level write lock on these related tables, because you don't want other
users to be changing data while referential integrity isn't being enforced.
(Alternatively, you can open the database in exclusive mode if the users
won't tar and feather you for blocking them from doing their work.) Drop
the foreign key constraints between these tables. Delete the record and
insert a new one with the same ID for the primary key. Repeat for as many
records as needed, then recreate the foreign key constraints again.

HTH.
Gunny

See http://www.QBuilt.com for all your database needs.
See http://www.Access.QBuilt.com for Microsoft Access tips and tutorials.
http://www.Access.QBuilt.com/html/expert_contributors2.html for contact
info.

 

[Keith Wilby]

Therefore, my question is this: Is there an easy way to detect which
field has been changed (without writing code behind each text box,
given that Access doesn't allow control arrays... It probably seems
like I'm being lazy, but with probably 8 or so forms, each with 15-20
text boxes, that would be a lot of work!)

I wrote a function to do this with any form for an audit trail, you just
pass the form to the function along with the record's unique ID. Here's an
extract from the code, you may be able to adapt it and probably make it more
elegant in the process:

Calling code:
Call libHistory(Me, Me.txtID)

Public Function libHistory(frmForm As Form, lngID As Long)

Dim ctl As Control

For Each ctl In frmForm
'Ignore controls such as labels
If ctl.Name Like "txt*" Or ctl.Name Like "cbo*" Or ctl.Name Like
"ogr*" Or ctl.Name Like "chk*" Then
'Record null to value, value to null, and value changes
If ((IsNull(ctl.OldValue) And Not IsNull(ctl.Value)) Or
(IsNull(ctl.Value) And Not IsNull(ctl.OldValue)) _
Or ctl.OldValue <> ctl.Value) Then
'DO STUFF
End If
End If
Next

End Function

Regards,
Keith.
www.keithwilby.com

 

[David W. Fenton]

Jet doesn't have deferred constraints like client/server databases
do, so you'll have to drop the constraints, not just "turn them
off."

I can't imagine a situation in which this would be advisable. What
happens if someone else adds non-conformant data to the database
during the time when the constraints have been turned off?

Changes to the schema should *never* happen incidentally as a part
of daily operation of an application. If the application depends on
that, then there's a horrendously bad design in place and that
should be fixed.

 

[David W. Fenton]

There are circumstances where you need to disable DRI, CHECK
constraints, etc temporarily while you complete a set of
operations.

I don't agree with this. I don't see any situations where RI should
be disabled.

 

[David W. Fenton]

Perhaps 'need' was the wrong word; substitute 'may find it
conducive'.

What about the situation where you need to alter a key's value and
can't use ON UPDATE CASADE on all you tables because there are
potential cycles or multiple update paths that the engine isn't
smart enough to resolve?

Then your schema is wrong.

Fix that and the problem goes away.

 

[Bri]

I wrote a function to do this with any form for an audit trail, you just
pass the form to the function along with the record's unique ID. Here's an
extract from the code, you may be able to adapt it and probably make it more
elegant in the process:

Calling code:
Call libHistory(Me, Me.txtID)

Public Function libHistory(frmForm As Form, lngID As Long)

Dim ctl As Control

For Each ctl In frmForm
'Ignore controls such as labels
If ctl.Name Like "txt*" Or ctl.Name Like "cbo*" Or ctl.Name Like
"ogr*" Or ctl.Name Like "chk*" Then
'Record null to value, value to null, and value changes
If ((IsNull(ctl.OldValue) And Not IsNull(ctl.Value)) Or
(IsNull(ctl.Value) And Not IsNull(ctl.OldValue)) _
Or ctl.OldValue <> ctl.Value) Then
'DO STUFF
End If
End If
Next

End Function

Regards,
Keith.
www.keithwilby.com

Great idea, exept that OldValue only works with Bound controls which
implies Bound Form and this is an Unbound Form. I use something very
similar myself in my Audit routine.

 

[Bri]

Perhaps 'need' was the wrong word; substitute 'may find it conducive'.

What about the situation where you need to alter a key's value and
can't use ON UPDATE CASADE on all you tables because there are
potential cycles or multiple update paths that the engine isn't smart
enough to resolve?

Jamie.

I assuming (hoping) that you are referring to a one time maintenance
issue vs a regular process. In this case you would have kicked everyone
out of the db, so as long as everything is back in place before you let
them back in then, yes, you might want to do this.

I agree with David that you do NOT want to have constraints modified in
an ongoing process. Its better to define constrains that suit your needs
and then leave them alone. The OP has been shown several alternatives
(Terry's function idea is the one I would use), so he shouldn't need to
consider the constraint option further.

 

['69 Camaro]

Hi, David.

I can't imagine a situation in which this would be advisable.

For data migrations, legacy data often doesn't conform to the structure of
the new tables. Client/server databases offer the ability to enable and
disable constraints, create deferrable constraints, defer constraints
initially, and validate/not validate existing data when the constraints are
enabled. In general, it's more efficient to use deferred/non-validated
constraints to allow non-conforming data to be imported into tables, because
fixing the data before importing into the tables is often more cumbersome
and time-consuming without the database engine to manipulate large data
sets.

What
happens if someone else adds non-conformant data to the database
during the time when the constraints have been turned off?

As I mentioned in my previous post, special precautions need to be taken:
"create a table level write lock on these related tables, because you don't
want other users to be changing data while referential integrity isn't being
enforced." With a write lock, the users can't alter the data in these
locked tables. They can only read the data. And if Andrew forgets to lock
the tables before beginning these operations, he may find that he can't add
the constraints to the tables again when he's done because the users have
added data that doesn't comply with the constraints. Andrew won't forget
the table locks more than once if he takes this approach.

Changes to the schema should *never* happen incidentally as a part
of daily operation of an application.

Agreed. This is a DBA function for database maintenance, not something
that's executed as part of the normal business logic of the application.

If the application depends on
that, then there's a horrendously bad design in place and that
should be fixed.

Of course it's a bad design for a normal user application. However, if it's
a DBA tool for migrating data where no users are in the database yet, then
it's not half as bad as it sounds, because it's a one-time (or occasional)
utility function. But I think it's much more likely that Andrew's approach
is intended for the actual day-to-day operations, not a one-time data
migration.

As for why I gave the instructions I did, Andrew asked whether or not it
could be done, so I provided a response in that context. I can give my
advice that this approach is ill-advised along with instructions on how to
do it, because I know that:

1. Some people will consider the steps outlined and, based upon their
experience, see that this approach is not in their best interest for daily
operations and therefore consider the other, more viable, options.

2. Some people will ignore the advice that this is ill-advised since
they're focused on the ultimate outcome, not the wisest path to reach that
outcome. Those are the ones that will implement the instructions and get to
see the consequences first hand. Depending upon how much experience they
have with relational database applications and how much traffic the database
gets, the consequences might not be evident to them immediately, but they'll
eventually see that this approach is not in their best interest.

And while I'd rather people not make mistakes when it comes to working with
data, I believe that we need to allow people to make small mistakes so that
they can learn from them, which will later help them to figure out how to
avoid many of the collosal mistakes. I know that when I was new to
relational databases, the experts who trained me let me make mistakes,
because they were confident that they could fix anything that I screwed up,
and I'd learn valuable troubleshooting skills along the way. Like most
people, I made many mistakes, but seeing the results of "this is why we
_never_ do such-and-such" hammered home the lessons I learned. I got to
learn how to fix those mistakes myself, so I never need to rely on others to
notice -- and fix -- my mistakes for me. I don't want to prevent others
from gaining this valuable experience.

HTH.
Gunny

See http://www.QBuilt.com for all your database needs.
See http://www.Access.QBuilt.com for Microsoft Access tips and tutorials.
http://www.Access.QBuilt.com/html/expert_contributors2.html for contact
info.

 

[David W. Fenton]

For data migrations, . . .

That's a one-time operation. The question was about having it be a
regular thing to turn it off and turn it on in order to input
certain kinds of data. That was also what someone else suggested.

There is simply on situation that I can think of where turning off
RI and then turning it back on should be a regular part of the
process.

If you need to regularly import data that needs to be massaged to
remove violations of your RI, then you need temp tables to operate
on it before it actually gets imported into the live database.

. . . legacy data often doesn't conform to the structure of
the new tables. Client/server databases offer the ability to
enable and disable constraints, create deferrable constraints,
defer constraints initially, and validate/not validate existing
data when the constraints are enabled. In general, it's more
efficient to use deferred/non-validated constraints to allow
non-conforming data to be imported into tables, because fixing the
data before importing into the tables is often more cumbersome and
time-consuming without the database engine to manipulate large
data sets.

I don't see it. Import buffer tables are the way to go, seems to me.

In any event, I can certainly see doing it once during a massive
import process. I can't see doing it ever for a live database.

As I mentioned in my previous post, special precautions need to be
taken: "create a table level write lock on these related tables,
because you don't want other users to be changing data while
referential integrity isn't being enforced." With a write lock,
the users can't alter the data in these locked tables. They can
only read the data. And if Andrew forgets to lock the tables
before beginning these operations, he may find that he can't add
the constraints to the tables again when he's done because the
users have added data that doesn't comply with the constraints.
Andrew won't forget the table locks more than once if he takes
this approach.

This all sounds completely insane to me. I would never pay a
developer who produced such a monstrosity.

Agreed. This is a DBA function for database maintenance, not
something that's executed as part of the normal business logic of
the application.

That's not what the original question was about, though.

Of course it's a bad design for a normal user application.
However, if it's a DBA tool for migrating data where no users are
in the database yet, then it's not half as bad as it sounds,
because it's a one-time (or occasional) utility function. But I
think it's much more likely that Andrew's approach is intended for
the actual day-to-day operations, not a one-time data migration.

I was responding to the question asked, which was about a regularly
used procedure, not a one-time import. I don't consider a one-time
removal of RI to import certain data to be part of the regular
operation of a database. It's an operation done before the schems is
finished and initialized with data.

As for why I gave the instructions I did, Andrew asked whether or
not it could be done, so I provided a response in that context. I
can give my advice that this approach is ill-advised along with
instructions on how to do it, because I know that:

1. Some people will consider the steps outlined and, based upon
their experience, see that this approach is not in their best
interest for daily operations and therefore consider the other,
more viable, options.

2. Some people will ignore the advice that this is ill-advised
since they're focused on the ultimate outcome, not the wisest path
to reach that outcome. Those are the ones that will implement the
instructions and get to see the consequences first hand.
Depending upon how much experience they have with relational
database applications and how much traffic the database gets, the
consequences might not be evident to them immediately, but they'll
eventually see that this approach is not in their best interest.

I don't see why you'd give the advice, given that you knew the
contemplated scenario was one that you'd never recommend.

And while I'd rather people not make mistakes when it comes to
working with data, I believe that we need to allow people to make
small mistakes so that they can learn from them, . . .

This is no small mistake, in my opinion. It's a fundamental error in
the understanding of how the schema should be established and how
data should be entered into that schema.

. . . which will later help them to figure out how to
avoid many of the collosal mistakes. I know that when I was new
to relational databases, the experts who trained me let me make
mistakes, because they were confident that they could fix anything
that I screwed up, and I'd learn valuable troubleshooting skills
along the way. Like most people, I made many mistakes, but seeing
the results of "this is why we _never_ do such-and-such" hammered
home the lessons I learned. I got to learn how to fix those
mistakes myself, so I never need to rely on others to notice --
and fix -- my mistakes for me. I don't want to prevent others
from gaining this valuable experience.

I don't see why one would answer a question as though a different
question had been asked, which is what it seems to me that you did.

 

[Keith Wilby]

Great idea, exept that OldValue only works with Bound controls which
implies Bound Form and this is an Unbound Form.

Hence the phrase "you may be able to adapt it".

Keith.

 

[Andrew R]

In reply to all those of you who replied - thank you. I didn't realise
that it would stir quite such a vigorous debate!

It's given me some things to think about, and to those who offered more
advice, I may get in touch again in the next few days.

Incidentally, a note to Gunny - no, that's not what I'm paid for! I'm
developing this db for my own needs and to further my knowledge of and
interest in Access. I'm actually an IT trainer, rather than a database
developer....

Thanks again for all your thoughts and contributions....

Regards

Andrew

 

[Andrew R]

Oh, and incidentally, the database is only for my own use in managing
clients and projects, so although it is fairly complex, the issues
which arise in a multi-user environment are not so relevant to this
particular case (although obviously they *do* impinge on the quality of
the design as a whole).

Thanks

Andrew

 

[David W. Fenton]

What about where the schema is 'right', in that logically all the
cascade paths can be resolved, but the engine isn't smart enough
to work this out. There's a live in another of the Access
newsgroups:

http://groups.google.com/group/microsoft.public.access.tablesdbdesi
gn/browse_frm/thread/b806c09e1a9cc11e/

Sounds like an adjustment to the schema is needed. I've had circular
relationships in Jet databases (even replicated) and have never had
a problem. Perhaps one solution is changing the FK to allow Null and
not be required is all that's needed, since I've never had a
circular relationship with a required FK value in the chain.

Of course, I also don't believe in cascading updates, since I am
philosophically completely opposed to using PKs that are ever
updated (I'm against natural keys in all but the most trivial cases,
such as lookup tables).

Either of those alterations to the schema should make it work.

You earlier comment was probably more correct in that you are not
seeing the situation (how can you tell me my schema is wong when I
haven't posted one <g>?!)

If the schema necessitates regularly turning off RI, then it's
wrong. I don't need to know anything else about it.

 

[Bri]

Hence the phrase "you may be able to adapt it".

Keith.

Well, you gave the function in response to the issue of dealing with
controls that had changed their value and your function uses OldValue to
do that... so, what is left to adapt?

 

[David W. Fenton]

We're fortunate in Access/Jet that the engine is quite good at
resolving cascade paths. Much better, for example, than SQL Sever:
as soon as it encounters two paths it chokes i.e. the classic
adjacency list, a popular common/popular design choice for trees
in SQL, cannot support cascading updates in SQL Server.

Well, I think cascading updates are a bad thing, as an updatable PK
is a bad thing.

Secondly, every N:1 relationship can be replaced by an intermediate
join table. It's not quite as intuitive as your usual N:N join table
(it's a degenerate case of the N:N), but it works just the same.

The point is that there are ways to avoid the circular relationship
problem by redesigning your schema. It will still reflect the
entities being modelled. It just do it in a different fashion.

It would be a complex design in Access/Jet that would cause the
engine to choke and therefore there would be a lot of scope for
alternative design choices. The more complex the design the more
difficult it is to say for sure that the design is 'wrong' because
of the many design choices and compromises between logical model
and physical implementation which inevitably have been made along
the way.

Cascading updates indicates to me that somebody has gone badly wrong
in the first place.

Temporarily suspending DRI in Access/Jet is not desirable (e.g.
can I create a serialized transaction that would prevent others
making schema changes and allow me to rollback mine? I really
don't know) and I've never had to do it myself but I stop short of
saying that doing so is always wrong.

It's bloody stupid to have RI in place that has to be suspended in
order to insert data.

Some people choose to use natural keys in DRI. Are you saying they
are 'wrong'?

Yep. Any data that can be edited shouldn't be used as a PK, because
the function of the PK is to relate data, not to store information
about the entities represented in the table.

That doesn't mean that unique indexes shouldn't be maintained on the
natural key (which may be composite), but natural keys lead to just
this kind of problem.

The only exception I would make is for one-column lookup tables,
where there is no dependent data (i.e., the PK is the entire data
for the entity).

I've having trouble envisaging what you mean. Could you post your
schema and some test data, please.

I posted mine (or rather, that OP's) in the other thread. Perhaps
you could also post a 'fix' to that schema too?

I've explained at least three different ways to avoid circular
schema problems. None of them is complicated enough to need me to
draw you a picture.

 

[Keith]

Well, you gave the function in response to the issue of dealing with
controls that had changed their value and your function uses OldValue to
do that... so, what is left to adapt?

You know, there will always be smart-arsed know-it-alls on these groups
who use their no doubt superior knowledge on a subject to try to
humiliate and score cheap points from other contributors. You fall into
this category. I always try to help other Access users where I can but
my contributions can sometimes be a bit off the mark and I'm happy and
grateful to accept corrections where necessary from those with better
knowledge, this is how I learn from my mistakes, but when those
corrections are offered in a holier-than-thou manner it really pisses me
off.

I am not so geek-like that I know about every method, every function,
every facet of Access. I am also sometimes guilty of reading posts too
quickly. In this instance, although I read the "unbound" part I failed
to digest it fully which is why I offered the code for adaptation and
not as a belt and braces solution. So I made a mistake, so what? I'm
sure the OP realised that without your cheap shots.

I read your exchange with David Fenton in the "Created on Access 2003,
but......................." thread with interest. David is extremely
knowledgeable in Access but, IMO, his bedside manner can sometimes leave
a little to be desired, but it pales into insignificance compared to
yours. This is not something I would be proud of if I were you, which
thankfully I am not.

<PLONK>

 

[Bri]

You know, there will always be smart-arsed know-it-alls on these groups
who use their no doubt superior knowledge on a subject to try to
humiliate and score cheap points from other contributors. You fall into
this category. I always try to help other Access users where I can but
my contributions can sometimes be a bit off the mark and I'm happy and
grateful to accept corrections where necessary from those with better
knowledge, this is how I learn from my mistakes, but when those
corrections are offered in a holier-than-thou manner it really pisses me
off.

I am not so geek-like that I know about every method, every function,
every facet of Access. I am also sometimes guilty of reading posts too
quickly. In this instance, although I read the "unbound" part I failed
to digest it fully which is why I offered the code for adaptation and
not as a belt and braces solution. So I made a mistake, so what? I'm
sure the OP realised that without your cheap shots.

I read your exchange with David Fenton in the "Created on Access 2003,
but......................." thread with interest. David is extremely
knowledgeable in Access but, IMO, his bedside manner can sometimes leave
a little to be desired, but it pales into insignificance compared to
yours. This is not something I would be proud of if I were you, which
thankfully I am not.

<PLONK>

Whoa, now!! Back up the truck. I'm sorry if you feel that my reply was
an attempt to be a smart ass and to score cheap points at your expense.
I can assure you that that was not my intent. I really was trying to
point out that your solution wouldn't work and why it wouldn't for the
benefit of the OP primarily and for you as well. You then replied that
the OP was to adapt your code for his needs. I then responded that after
you remove the OldValue part of the function that there was nothing left
that could be used to solve his problem as the OldValue was the key part
of that function. I was wondering if, after the OldValue part of the
function was dismissed, what you thought was still there to adapt? It
was a question, in case I had missed something else in there. I'm sorry
you took offense to that. Perhaps I could have written it differently.

I'm not sure what there is about my responses that 'pales into
insignificance'. Perhaps, you would explain to me the error of my ways?
I certainly don't think of myself as hard to get along with, but perhaps
there is something in the way I write that comes off that way. I was
unaware of it. What could/should I have done differently?

In the exchange with David you refer to, he refused to acknowledge that
what I said I had done was true, he said I was a lier, and even when
faced with several sources of reference to back my story up, he still
refused to believe it. I can't see how you would place me as the
aggressor in that thread.