Hello

I found that price of += operator on string is too high in jython. For
example 5000 such operations took 90 seconds (i generated html copy of
table with 1000 rows and 5 columns). Generation of row data into separate
string and joining after lead to time 13 seconds !!!

What's alternative way to do that ? (similiar parts of my code are terribbly
slow and such simple solution as above didn't help).


Thanks,
Jan

> I found that price of += operator on string is too high in jython. For[color=blue]
> example 5000 such operations took 90 seconds (i generated html copy of
> table with 1000 rows and 5 columns). Generation of row data into separate
> string and joining after lead to time 13 seconds !!![/color]

Its generally not recommended to use simple string concatenation for
building larger strings - neither in python nor in java/jython.

afaik there are two solutions to this: The java-way and the jython way:

java: Use StringBuffer
python: use a list, and append the strings to that. Then, when you want the
result, do

"".join(stringlist)

--
Regards,

Diez B. Roggisch

Jan Gregor wrote:[color=blue]
> Hello
>
> I found that price of += operator on string is too high in jython. For
> example 5000 such operations took 90 seconds (i generated html copy of
> table with 1000 rows and 5 columns). Generation of row data into separate
> string and joining after lead to time 13 seconds !!!
>
> What's alternative way to do that ? (similiar parts of my code are terribbly
> slow and such simple solution as above didn't help).[/color]

I don't use Jython, but are you not able to do something like:

string_list = []
for ... in ...:
...
string_list.append(...)
...
string = ''.join(string_list)

This is the usual Python idiom and is usually faster than the += idiom.

Note too that +ing strings in Java also has this problem -- hence
StringBuffer or whatever it's called.

Steve

Ok, thanks. I didn't think that += operator is nondestructive operation
- but strings are immutable so this makes sense.

On 2004-12-13, Diez B. Roggisch <deetsNOSPAM@web.de> wrote:[color=blue][color=green]
>> I found that price of += operator on string is too high in jython. For
>> example 5000 such operations took 90 seconds (i generated html copy of
>> table with 1000 rows and 5 columns). Generation of row data into separate
>> string and joining after lead to time 13 seconds !!![/color]
>
> Its generally not recommended to use simple string concatenation for
> building larger strings - neither in python nor in java/jython.
>
> afaik there are two solutions to this: The java-way and the jython way:
>
> java: Use StringBuffer
> python: use a list, and append the strings to that. Then, when you want the
> result, do
>
> "".join(stringlist)
>[/color]

StringBuffer class from java was right solution - yours looses encoding,
and in jython I was unable to get it back - in python it worked fine.

Jan
[color=blue]
> I don't use Jython, but are you not able to do something like:
>
> string_list = []
> for ... in ...:
> ...
> string_list.append(...)
> ...
> string = ''.join(string_list)
>
> This is the usual Python idiom and is usually faster than the += idiom.
>
> Note too that +ing strings in Java also has this problem -- hence
> StringBuffer or whatever it's called.
>
> Steve[/color]

Jan Gregor wrote:[color=blue]
> StringBuffer class from java was right solution - yours looses encoding,
> and in jython I was unable to get it back - in python it worked fine.[/color]

If you mean that Jython returned a string, when the inputs were unicode, then
that can probably be fixed with:

result = u''.join(string_list)

(Python switches to the unicode version automatically if it finds any unicode
strings in the supplied sequence. Jython may not do that - I'm not a Jython user
though, so I'm not sure).

Cheers,
Nick.

--
Nick Coghlan | ncoghlan@email.com | Brisbane, Australia
---------------------------------------------------------------
http://boredomandlaziness.skystorm.net