[AD] If I don't put this here, someone might complain...

Our tools for going MV to SQL do exactly that.  We have a mapping wizard that 
analyzes the dictionaries and the data.  Depending on how good the dictionaries 
are, we get close to 80% right first try.  Then you get to test and tweak, all 
in a graphical environment, until you have it 100% right.  The tool lets you 
even decide what %age of bad data you'll ignore.  It's not uncommon to have a 
handful of bad records in a big file.  Set the tolerance to 1% and if you have 
a dictionary that says attribute 2 is a date, provided less than 1 in 100 items 
have something that's not a valid date (you can set the range of values that 
work for your system, too) we'll call it a date and use that dictionary for the 
field name.

We also have tools that let you identify bad records so you can actually find 
and fix them.

The map is ultimately stored in the file's dictionary.

I think we've gotten off-topic, again, though. The original question was about 
going the other way.

Going SQL to MV, Item-id selection is likely to be the biggest hurdle.  You do 
not need to set a primary key in SQL Server.  Even if you do, you may have a 
unique index in SQL which would make a better candidate for the item-id.  A 
tool to migrate can make a best first guess, but it will sometimes get it 
wrong.  At some point, someone familiar with the application, with an 
architectural mindset, is going to have to look at the output and probably 
tweak the results.  I just don't see a substitute for this.

On the other hand, if you just get the data over, someone can convert it 
locally to whatever format they ultimately want.  Our tools typically exist to 
help get the data to where the application programmers are, and they take it 
from there.   Then the application guys figure out what to do with it from 
there.  If you have to repeat this from time to time, you use a staging 
approach; you pass the raw converted data to a temporary file, then the 
application people process this file and convert it to their ultimate format.

So, you get to choose between doing it all on the first pass, or getting the 
data over and letting someone at the target end (in this case, MV) complete the 
transformation.  You are choosing between "E -> TL" and "ET -> L".  Where T 
happens is often driven by where the programming muscle in the project resides.

-----Original Message-----
From: [email protected] 
[mailto:[email protected]] On Behalf Of Marc Harbeson
Sent: Tuesday, December 28, 2010 8:13 AM
To: 'U2 Users List'
Subject: Re: [U2] Migration

LOL

In my mind - there would be a operator map tool here - I don't think the tool 
could be self aware enough to figure out every possible combination of 
everything.

It could certainly "guess" 80% correctly and be corrected on the remaining 20% 
on suggesting maps.

I see this with data going from MV to SQL - sometimes the dictionaries are just 
wrong...  it's easier to adjust using the map...  (in particular if they're not 
your dictionaries to adjust)

I would imagine an adjustment could be made to the maps to increase MV 
performance - just like you would in SQL when porting MV data there.

-----Original Message-----
From: [email protected]
[mailto:[email protected]] On Behalf Of Robert Houben
Sent: Friday, December 24, 2010 1:41 AM
To: U2 Users List
Subject: Re: [U2] Migration

Oh, one more point.  What if your SQL environment had NOT defined a primary key 
for APPOINTMENTS, but had multiple indexes, one of which happened to have 
CUSTOMERNO, APPTDATE, APPTTIME and APPTTYPE.  How would you figure out what to 
use as the item-id of the PICK file?

What if you had a SQL table that actually did not have a set of fields that 
guaranteed a unique value? Then you have NOTHING to create an item-id from!

I have to stop this, it will consume me! :o  But the list goes on.  Oh the 
humanity!


_______________________________________________
U2-Users mailing list
[email protected]
http://listserver.u2ug.org/mailman/listinfo/u2-users
_______________________________________________
U2-Users mailing list
[email protected]
http://listserver.u2ug.org/mailman/listinfo/u2-users

Reply via email to