|
Mycroft Holmes wrote: For one, I'm told Oracle does not like multiple databases on a server, the design is to use schemas, I thought this was pure bullsh*t but it came from a DBA so it may be right.
It's BS. It works quite fine. It's just pointless. Multiple schemas is the way to go.
Mycroft Holmes wrote: For another Oracle is case sensitive when dealing with data so where name = 'Johny' will miss 'johny' IMHO this and the ongoing support requirement enough reason to use another database (sql server).
And the problem is what? Use a function based index and you got the best of both worlds.
Example:
CREATE INDEX foo_bar_ix
ON foo (
LOWER("bar")
)
/
And then you can search the table with Select * from Foo where lower(bar) = lower('Johnny');
"When did ignorance become a point of view" - Dilbert
|
|
|
|
|
Jörgen Andersson wrote: Use a function based index and you got the best of both worlds
Nice. I didn't know you could do that. A 5 just for that trick.
Jörgen Andersson wrote: It's BS. It works quite fine. It's just pointless
I suspect he's referring to the load on having multiple instances running as opposed to a single instance running multiple schema.
|
|
|
|
|
Pete O'Hanlon wrote: Nice. I didn't know you could do that. A 5 just for that trick.
That single trick alone is worth the extra trouble of using Oracle
"When did ignorance become a point of view" - Dilbert
|
|
|
|
|
Materialised Views. They are worth it as well. Mind you, we have a top notch Oracle DBA to make sure that I don't have to know how to do this.
|
|
|
|
|
Pete O'Hanlon wrote: Materialised Views. They are worth it as well
Agreed.
But sqlserver has indexed views.
Pete O'Hanlon wrote: Mind you, we have a top notch Oracle DBA to make sure that I don't have to know how to do this.
I guess that's the major problem with Oracle, that you need one.
"When did ignorance become a point of view" - Dilbert
|
|
|
|
|
My biggest problem with materialised views is that we can't use any SDO geometry types in them (something that would be really useful for us).
|
|
|
|
|
Are you sure about that?
I understand that it's severely limited[^]. But not impossible.
"When did ignorance become a point of view" - Dilbert
|
|
|
|
|
It's point 9 that's the killer. We have some data that comes in as x,y coords and we want our materialised views to be spatial, so you get screwed up when you try to construct a location sdo geometry.
|
|
|
|
|
Spatial is totally out of my line of work, but I would have thought point one or ten would have been the real killer.
I'm curious, what would have been the purpose of the mv if it had been possible?
"When did ignorance become a point of view" - Dilbert
|
|
|
|
|
We are reliant on a separate supplier for a standardised gazetteer implementation in one particular project (the data's provided by the government), but we need it in a different format (a spatial one for spatial searches). We need the view to be updated on import of data into the master gazetteer. We've worked around the issue, but it would be nice not to have to work around the issue.
|
|
|
|
|
I've been thinking about this one for a while.
If I understand you correctly (and it's quite possible I don't), you're getting XY coords from an external source that you load into a table, and you want them transformed into a sdo_geometry datatype which you do via a materialized view. And for some reason you can't do this at the import.
One possible solution is to add a virtual column into the master gazetteer that calculates the sdo_geometry.
From 11g you can index, gather statistics and partition by a virtual column, which makes them a lot faster and quite useful nowadays.
There are probably other limitations that makes it impossible to drop the view, but nothing I could find on a quick google.
"When did ignorance become a point of view" - Dilbert
|
|
|
|
|
Pete O'Hanlon wrote: we have a top notch Oracle DBA to make sure that I don't have to know how to do this.
I wish. Difficult to find and expensive to retain. I'm in the throes of getting to know Oracle again after more than a decade in SS, I think SS must be like VB, lots of work to make it easy for the developer.
Then I aggregated 150m records in minutes where SS was taking hours and remembered why we are using Oracle.
Never underestimate the power of human stupidity
RAH
|
|
|
|
|
I've balanced out the univote.
|
|
|
|
|
Oh, hadn't noticed. Thanks!
"When did ignorance become a point of view" - Dilbert
|
|
|
|
|
Jörgen Andersson wrote: It's BS. It works quite fine. It's just pointless. Multiple schemas is the way to go.
I suspected as much, it was probably a DBA enforcing the schema requirement.
Jörgen Andersson wrote: And then you can search the table with Select * from Foo where lower(bar) = lower('Johnny');
I will need to do some work to understand the relevance on the function based index. However I'm used to SS where case is irrelevant when comparing text, there for the lower() is not required. This will continue to bite me until I get used to it.
My real peeve with Oracle is the all upper case objects, I hate underscores so my names look like FILENAMEDSOMETHING instead of FileNamedSomeThing.
Never underestimate the power of human stupidity
RAH
|
|
|
|
|
Mycroft Holmes wrote: I will need to do some work to understand the relevance on the function based index
Say for example that you have a log table where one of the columns has the date datatype which carries both date and time, and you want to get all occurences from one specific day.
A query for trunc(logdate) would make a full table scan, unless you have an index on trunc(logdate).
Another example, You have a really large table where you have a status column, and the only value you ever make a search on is 'PENDING'. Then an index on "Case When status = 'PENDING' Then 1 Else Null" will be very small and fast, as only those entries where the status is PENDING will be stored in the index as null values are not indexed.
The backside is that function based indexes is costing more to maintain.
"When did ignorance become a point of view" - Dilbert
|
|
|
|
|
Thanks for that Jorgen
While the ability to do these tweaks is very good, the requirement to do them is painful. I beging to understand why a full time DBA is a requirement, I consider this type of tuning beyond the requirements of a developer.
Never underestimate the power of human stupidity
RAH
|
|
|
|
|
Mycroft Holmes wrote: , the requirement to do them is painful
Having to use Oracle?
Mycroft Holmes wrote: beyond the requirements of a developer
The learning threshold is indeed high, especially if you come from the world of SqlServer.
BTW, I liked your comparison with VB, had it been more catchy I would have stolen it for a sig 
"When did ignorance become a point of view" - Dilbert
|
|
|
|
|
V. wrote: What do I need to watch out for when designing a new database ?
I'm with Mycroft, try to save a dime here and it'll cost a fortune down the road.
V. wrote: but we do have some knowledge about Oracle
I have some knowledge about cars, but that doesn't qualify me to design a new one. People that are taking design-decisions should bring forth rational arguments for their proposals.
I are Troll 
|
|
|
|
|
What's the reasoning behind getting several databases?
If it's getting slow because of large dataamounts you should have a look at partitioning instead.
Get a good DBA for the project to give you proper advice. It'll pay back later.
"When did ignorance become a point of view" - Dilbert
|
|
|
|
|
Jörgen Andersson wrote: What's the reasoning behind getting several databases?
No idea, it blew my mind as well. we still need to go for our first meeting, so if it is up to me this idea will go straight through the window.
V.
|
|
|
|
|
V. wrote: someone opted for multiple smaller databases
sure, why keep things simple when it is so easy to make them complex?
you could as well store each data item in a separate file...
|
|
|
|
|
Luc Pattyn wrote: you could as well store each data item in a separate file...
Of course! That's it. Let's save everything in a nice (bloated) XML file complete with metadata, triggers, constraints, stored procedures, the works ... see how that works 
V.
|
|
|
|
|
V. wrote: nice (bloated) XML
Know a guy who said XML was the next thing for storing data, a year later he deigned he said it.
|
|
|
|
|
I know a company who used xml as the format for etl between major banking systems, stupidest decision they ever made, it worked well with the test system then we introduced them to a production size file.
Never underestimate the power of human stupidity
RAH
|
|
|
|
General 
News 
Suggestion 
Question 
Bug 
Answer 
Joke 
Praise 
Rant 
Admin 
Submit an article or tip
