Question about efficiency, should i use hashmap?

Question

Hi all
i read a csv file and store each line's data into a class MyData.
this class contains a variable of id - unique for object.
i want to store all my MyData object in a collection for later use.
if i want to access them in the fastest way, should i store my objects in a hashmap, and the key value would be their id?
like -
HashMap hashmap = new HashMap();
MyClass mc1 = new MyClass(1); //id =1
MyClass mc2 = new MyClass(4); //id =4
//i add the objects
hashmap.put(mc1.getID(),mc1);
hashmap,put(mc2.getID(),mc2);
//then when i want to get a certain object, where id is 4:
MyClass mc = hashmap.get(4);should i use other collection? or is it fine that way?
Thanks!

Answer 1

jwenting wrote:
don't worry about microoptimisations like selecting a collection class based on how fast it is unless and until you have written evidence that the class you had initially chosen was the sole reason for unacceptably poor performance (which will hardly ever matter, the only time I've ever encountered that was with an ancient web application making extensive use of Vector, where switching to ArrayList made it an order of magnitude faster).Hmmm... I do diagree with that statement, but then again I'm an old f@rt... I cut my teeth on the gory gizzards of all manor of data-structures, and I still believe that selection of the "correct" data structure(s) is a core implementation issue... especially where efficiency is (even occasionally) desirable.
I haven't seen the same big gains when replacing Vector with ArrayList... in fact, I've only seen a very marginal (<10%) increase in "overall" throughput... I have seen a couple of orders of magnitude (or thereabouts) increase when I swapped a Vector<Integer> for an int[], but that was in the inner-loop of an inherently n^2 (I think) spatial algorithm.... and I also made it reuse the int[], instead of creating, growing and throwing away the vector every time through the outer loop. Malloc is expensive ;-)
I'd be interested to hear more details about under the conditions under which ArrayList is 10 times faster than Vector, because I do a lot of work on a 1.3/1.4 era codebase, which is full of Vectors and HashTables... and I would like to be able to make a case for an en mass replacement (and generifying) of the "old school" datastructures... but alas, I see very little bang for our buck, so I just bite my toungue and continue using Vectors, even in new code, because so many existing "helpers" take a (explicitly) Vector parameter.
I heard a "rumor" somewhere (probably a thread here) that there's talk of discontinueing support for non-generic collections in 1.7, because they (the API/VM developers) are having to (sort of) maintain two divergent versions of every collection related thing... and nothing annoys a developer more than that, right... so Please can anyone confirm or repudiate this "rumour".
Cheers. Keith.