Re: Search Algorithm

From: Marc Rochkind (rochkind_at_basepath.com)
Date: 08/15/03 

Date: Fri, 15 Aug 2003 14:19:26 -0600

On Fri, 15 Aug 2003 20:09:27 GMT, Fletcher Glenn <f-g-l-e-n-n@quest.com>
wrote:

> Marc Rochkind wrote:
>>
>> On Fri, 15 Aug 2003 08:40:36 -0700, John Smith <John_Smith273@cox.net>
>> wrote:
>>
>> [snip]
>>
>> > I was considering hashes but they are not suitable for my
>> > application since my function would be called from a non-C language.
>> >
>>
>> I don't understand this... can you explain?
>>
>> How many entries are there? How critical is the search time? Without
>> this
>> info, it's impossible to know what's best. If the time is not critical,
>> it
>> is a waste of time to even ask the question. If the entries are few, you
>> can't beat linear search.
>>
>> Generally, the approach is:
>>
>> 1. Implement linear search.
>> 2. Profile the app.
>> 3. Reimplement the search when it rises to the top of the priority list
>> of
>> things to do, if ever.
>>
>> --Marc
>
> This is a rather naive assessment. If you know going in that the
> list could be very large, it is better to design efficiency in at
> the outset instead of ignoring a potential problem and hoping that
> you might someday have time to handle it.
>

Well, if you read "ignoring a potential problem" into my post, then I
haven't been clear. What should be done at the outset is to design in the
ability to change design decisions that might have to be changed. (A
principle of modular programming that predates OO by a decade or so, and
since made even easier with OO.)

Making those fined-grained, limited-scope design decisions should not be
done at the outset. Rather, the emphasis should be on correctness, clarity,
and speed of development. Runtime efficiency is at most fourth. That way
there's time to get the overall architecture right, and modules are
available earlier for integration and system testing rather than later.

The problem with early optimization is that development time and money get
exhausted during the initial development, since it's rare that the
information about where to optimize is available at that time.

So, on the whole, I would strongly disagree with your recommendation.

--Marc

Relevant Pages

  • Re: Search Algorithm
    ... >> list could be very large, it is better to design efficiency in at ... What should be done at the outset is to design in ... > are available earlier for integration and system testing rather than later. ... Performance and getting the architecture right are not ...
    (comp.unix.programmer)
  • RE: Unable to see entire content - Forms View
    ... If this does not work can you post the format of the form and also any On ... she can see an entire screen of her entries; she can then scroll ... Again, it looks fine in the Design View, ...
    (microsoft.public.access.forms)
  • New Michigan License Plate to be chosen this summer
    ... Winning design will be picked from more than 1,000 resident entries by panel of local officials. ... The next Michigan license plate could feature historic cars, images from the Great Lakes or one of the state's popular landmarks. ... Secretary of State Terri Lynn Land issued a challenge to residents less than a month ago to send in their own design ideas for the new plate, which replaces the state's No. 2 most popular license plate that shows the Mackinac Bridge. ...
    (misc.transport.road)
  • Re: What is this key?
    ... these entries remain in the registry. ... > I'm unaware of the specifics of the design here and couldn't speak to ... Was using the friggin web based reader ... >> rights. ...
    (microsoft.public.windowsmedia.player)
  • Vote for the best World Cup stamp design in our competition - and win a prize!
    ... The entries we've received in Positively Postal's "Design a World Cup Stamp" ... competition are now displayed on the site - and it's time for you to choose ... Colchester WebSite Sales is a Technology Consultancy that helps you get the ...
    (uk.rec.competitions)