Re: dataformatting in nodeJS SQL interface with IBM db2a

Henrik, I'll provide an inline response:

On Mon, Mar 26, 2018 at 11:42 AM, Henrik Rützou <hr@xxxxxxxxxxxx> wrote:

Nathan



well very clever RPG programmers has apparently decided to define all their
columns as CHAR fields instead of VARCHAR in their legacy database and that
will give technical inconsistency if connecters don’t follow the chosen
column design in the DB, so don’t shoot the messenger (the interface) that
does comply to the DB design but shoot the DB designer.

I agree with using VARCHAR fields. But that doesn't address the issue of
legacy databases.

I don't suggest shooting the messenger (the interface). Just replace it
with granular web service APIs.

When that is said it is a bad idea to have trailing blanks in a browser
based UI simply because users has to delete characters the can’t see in
order to be able to insert characters in the string otherwise it should
trigger an error because they will try to pass a 40 char long string into a
30 char long result field – what do you think happens in a SQL statement if
you try that?

Yes, I'm aware that trailing blanks are a disservice to browsers.

This not either a question on open access from outside but rather
controlled access between native and PASE and if there wasn’t that you can
forget of every thing that runs in PASE such as PHP, JAVA and all the Open
Source modules available.

Okay. But if you provide granular Web service APIs, they can be invoked
from PASE, as well as from external clients.

And even if there was some form of trimming in the interface (you can code
it in the SQL) you would still have the problem with legacy “numeric” date
fields, booleans thet sometimes are 0/1 and other time Y/N but never
unsupported true/false etc. that is spread all over the IBM I databases.

Okay. I'd suggest correcting those problems in code that runs in the same
address space as the database, as opposed to "in an application".

And in basics there are no difference in repacking/mapping raw DB table
data to either JSON, XML, X.12 etc. – as well as you have to do mapping of
the receivables or does SQL or DB supports raw FORMS encoded data or URL
gets escape sequenses - I hardly believe that - so you have to have a layer
where you do it.

You can obviously perform just about any type of logic in client
applications. But I still see advantages to running data mapping code in
the same address space as the DBMS.