| |
| |
| |
|
Page: 1 2 3 4 5 6 7
Comments:
<0> thanks
<0> bbl
<1> np
<2> I have a question about querry overhead, does anyone have any advice/rules of thumb, on when two seperate querrys pays off in the application layer, here is an illustartion.
<2> I have two tables one with (id,item) one with (id,manufacturer) i have x numbers in (id,item), and 1 in (id,manufacturer). Appx how high would x have to be to warrent the overhead of two seperate select statements
<2> 10,100,1000,10k,100k ? I know it is a rough question with alot of variables, just wondering when i should start benchmarking and when i know with certainty
<1> i don't get it... what query are you talking about?
<1> and why would 2 queries be more efficient than 1
<3> somebody msg me
<2> tenfour: Because 2 querryes would transmit less data
<2> querries*
<2> err damit
<4> how would it ever transmit less?
<2> if (id,item) in tbl1 has 10k rows, and (id,manufacturer) has 1 row, then after a join i would 10k rows+ 10krows, vs 10k rows + 1 row.
<1> eh that's not how it works at all
<4> thats when you group by and rollup
<5> say what?
<4> and only show the manufacturer once
<4> and save that data
<4> but you didnt just full scan twice
<1> also i think he's confusing joins & unions
<4> nor did you need to manually parse data later
<4> i think he's just stupid
<1> ecarroll_ if the data you want returned by the query stays the same, then there's no difference in transmission
<1> if you want 10k rows, then it's 10k rows whether you do 1 query or a million
<2> I think your just stupid twat, I have two tables tbl1 = (id,name) tbl2 = (id,make); I have 10k rows in tbl1, 1 row in tbl2, I can join and get 20k rows, or tranmit two selects with 10k+1 rows, the question pertains to a general rule of thumb.
<2> childish insult > rjarett
<1> you don't understand what a join does
<2> I already know the manufacturer number, but there is a one to one coralation to name:number, so i only really need one row of that. however I can join and replicate the name reducing it to one transmission
<1> ecarroll_ are you reading what i'm saying?
<4> the rule of thumb is, you are stupid
<1> if you need 1 row, then that's what will get transmitted
<1> whether you join, union, write 1 select or a thousand, or however you decide to get at it
<2> I think your failing to understand the scenario.
<2> I need all the rows.
<1> it's not like you would query for 10k rows and then filter out on the client
<2> the items are unique.
<2> I understand there.
<1> oh i think i see what you mean now... you don't want columns to be replicated 10k times
<1> i would use with rollup for this if there was a performance problem
<2> I have a table of items. with two columns. id and name. unique id to name, I need that total list, and then i link them through fkey to a manufacturer table, with (id,name) all items are made by the same manufacturer, should i marshell two select statmenets into one with a join, resulting in total rows in tbl1 * 2 columns (item names and manufacturer), vs two selects with total rows in tbl1 with one column (name) and then one querry on manufacturer with one
<2> tenfour: Exactly =D
<1> exactly? heh you are talking about 2 queries; i'm talking about with rollup
<1> which is it
<4> so how do you know which manufacturer did each item?
<1> read up on with rollup
<2> I'm wondering if i should run two queries to eliminate the replicated data, or if i should replicate the data to eliminate two trips to the server
<4> sicne you only have item_id, item_name.
<4> wheres yourt item_id, manuf_id lookup table?
<4> you redesign your app to be less stupid
<4> and what happens when identical items are made by different manufacturers?
<2> rjarett: Because this is a watered down scenario, I really get the manufacturer number, and have to link it to the items table to get all of the products made. but if i join them my result set would have alot of rows with item_name | manufacturer_name, such that the manufacturer_name is the same for each row
<4> you are a watered down thinker
<4> plain stupid
<2> I use a 1-to-many, for the join
<1> harsh :/
<4> learn to subselect brainiac
<2> Wtf does a subselect have to do with anything you ****ign spic.
<3> huh?
<1> ecarroll_ if your app is querying for a specific manufacturer, then it already knows the manufacturer
<1> so, why would you need that info anyway?
<2> tenfour: It knows the manufacturer number, not name.
<2> and with that number it can join the item table, that has an id, called fkey_manufacturer, I need only the name from the item table, and only the name of the manufacturer from the manufacturer table, I have lots of items
<1> ok i would probably always do 2 selects
<4> select a.item_name, b.manuf_name from item_table a, manuf_table b where item_id = (select item_id from item_manuf_lookup where manuf_id = ###)...
<1> because that's most likely what your app would be conducive to
<4> you can do it thousands of ways smarter than your idea you stupid idiot
<2> That isn't at all what i'm talking about you stupid nigger.
<4> natural join, basic left joins
<4> union all
<4> boy you sure are dumb
<2> I don't need left joins there nor is a union applicable. there is a one to one coralation.
<2> and I cant union because then i don't the difference between data from manufacturer table, and item table.
<4> if you were smart enough to know what you neam you wouldnt have asked a purely stupid question like "should I run 2 queries?)
<2> unless i make the ***umption that the appropriate row is at top/bottom
<4> what does top or bottom have to do with anything you moron?\
<4> there is no location in a database
<3> dang
<3> cant u guys discuss this without name calling
<2> union without order by standard puts rows in the order of the union.
<3> lIke NO U ****ING SPIC
<3> NO U ****ING MORON
<3> jeez
<4> Arrakis no this is more fun
<6> wow. The swedish police are looking into going from Oracle to MySQL. I wonder what kind of requirements they have of their databases ...
<4> he wont realize he is stupid unless he is told
<2> select 1 from foo union select 2 from bar; 1 is at top.
<0> hey why does MySQL dates look retarted?
<5> Inge: The ability for the data to corrupt randomly and accept bad data implicitly.
<2> I'm ignoring you now.
<4> wow its at the top huh?
<3> <6> wow. The swedish police are looking into going from Oracle to MySQL. I wonder what kind of requirements they have of their databases ...
<3> Inge- - they dont wanna be raped?
<3> they dont wanna be sued?
<6> heh
<3> they're sick and tired of Oracles business practices, licensing scheme or security flasws?
<6> would htink there must be some way to sue them for that.. :)
<3> could name a million things Oracle does wrong
<3> there probably is
<5> Sure, but all of them pale to the things that MySQL does wrong.
<3> but they probably cant get away with it
<4> they want to schedule court cases for 31-FEB-2006 and oracle wont let them
<3> caue only in america can you sue a company for not using their product :P
<6> ecarroll: If you insert 1 and then 2 into a table, and do a select on the table, odds are that the 1 will appear on top too...
<3> hahaha rjarett
<3> lmao
<3> :D
<6> hehehe
<3> I wouldnt be suprirsed if it was more politically motivated
<6> could be, could be
<3> politicks seems tob e moving alot of changes more so then common sense
<3> I'd expect sweden and europe to slowly move towards an all
<6> I don't like vendor lock-in, but I just don't think I want OpenOffice in preference of Office
<3> open source / GPLEd world
<5> Some dip**** wants to save money to look like a hero and is too clueless to know that pgres exists.
<3> Europes motive is to kill anything that makes money
<3> software makes money
<3> Halo_Four - i doubt pgsql could provide the support sweden required
<3> *would require
<7> for them its simple, their app handles the data security, they dont need the 'features' of a real rdbms
<7> thats the arguement you will get everytime
<3> like data corruption ?
<3> lol
<3> like seriously dude
<3> the latest mysql prrior to 5
<3> corrupted 3 times in a week
<3> on perfect hardware
<3> 3 column table
<3> 33333333 ****ing columns
<1> huhu mysql though :(
<3> i mean seroiusly 3 ****ing columns
<3> its gotta be the only db i know of that has that type of record
<1> i'd rather use the windows registry
<3> lmao
<3> oh well - best of luck to them :)
<4> ive had bitrot on 1 block on a 2tb database.. and oracle fixed it for me the next time i went to backup ;)
<4> i love oracle
<3> I'm biased against oracle
<3> mainly because of their prior tactics
<8> mm
<3> and their rapage licensing scheme which for the most part eliminates me from licensing them in any small business world
<3> there needs to be a 200-300 dollar Db
<3> :)
<3> i can pitch that and spring for that
<5> I think Oracle just changed their licensing plans for multiple core CPUs so that they only charge once per physical CPU.
<3> heh
<5> 200-300 DB? What are you selling?
<3> starting to catch up
<1> what active business can only afford $300 software?
<8> heh\
Return to
#sql
or
Go to some related
logs:
#beginner
russian fest wear
#politics
#stocks
windows genuine advangate
#computers
flash 8 +getfile
#firebird
#computers
ircn 8 wikipedia
|
|