[XML-SIG] I am confused...
Roman Suzi
rnd@onego.ru
Mon, 29 Jan 2001 22:10:09 +0300 (MSK)
On Mon, 29 Jan 2001, Martin v. Loewis wrote:
>> The above code is what I avoid to do. I want my application to be
>> completely data-driven, so even "/article/author/name" must not appear in
>> the program!
>
>I'll look into your code separately, but I'd like to make two points
>here:
>
>a) There is often a trade-off between data-driven and fast
> algorithms. Somebody will probably shoot me for that statement, but
> you should be willing to accept some performance degrading if you
> need it very general.
In C - yes, but in Python - I doubt.
Data-driven programs are shorter, contain less errors
and (IMHO) are faster.
>b) In Python, it is often possible to transform a data-driven approach
> in one with explicitly coded decisions, due to the dynamic nature
> of the language. If all else fails, you could generate the a program
> from the data.
This is true. But This add more complexity.
>c) I very much doubt that your *application* really needs to be
> completely data-driven; in any specific installation, there will be
> only a small set of queries. So that seems rather like a "nice to
> have" but a "must have" requirement.
I agree. I have this working now - but am not satisfied, because
do like to make changes in one place instead of hunting
them thruout many places.
My points are (they are drived by laziness ;-)
a) Software solution must be as general, as possible
(I think its a myth that less general solutions are harder,
longer to implement or are much less efficient:
2+2 is not easier than x+y, why hardcode x+x ?;-)
b) One parameter change requires one change in the code
("write everything once")
(if some nontrivial constant repeats in the code in the same role - its
a variable ;-)
c) Count total time of solution: time of programming
+ time of execution. (Not forgetting time of reprogramming!)
(In my case I better wait 3 more seconds than make
hell from supporting my solution)
Now I am turning toward XML & co. because it happen to
be a common data model to store such data I have for
web-site. Anything else is reinventing the wheel.
However, I want to apply the same design principles
(expressed above) while dealing with XML.
>Well, that's three points :-)
I think this branch of discussion is kinda offtopic.
Probably one day I will write a test for programmers
where there will be questions like:
#. What do you prefer more:
a)
if a == "1":
b = "5"
elif a == "4":
b = "20"
# ...
else:
b = "5000"
b)
b = {"1":"5", "4":"20", ..., "1000":"5000"}[a]
c)
b = str(int(a)*5)
d)
try:
b = str(int(a)*5)
except:
b = "5000"
:-)
For now my answer is (d) but there are cases where (c or d)
are not possible - then it will be (b).
>Regards,
>Martin
Sincerely yours, Roman Suzi
--
_/ Russia _/ Karelia _/ Petrozavodsk _/ rnd@onego.ru _/
_/ Monday, January 29, 2001 _/ Powered by Linux RedHat 6.2 _/
_/ "The tuna doesn't taste the same without the dolphin." _/