Here is a simple class to convert a string in from "1234.4567" into a
floating point.
It doesn't support hexadecimal or 1E23 notations but it is here to
show the idea
===================================
public class String2Float {
final int max_exp = 38;
final int max_str_len = 20;
float [] values;
char [] buffer;
public String2Float() {
buffer = new char [max_str_len + 1];
values = new float [max_exp + 1];
values[ 0 ] = 1;
for( int i = 1; i < max_exp; i++ ) {
values[ i ] = values[ i - 1 ] * 10;
}
}
public float Value(String str) {
int len = str.length();
if( len > max_str_len )
len = max_str_len; // ToDo: add proper error handling
// extract all characters into the local buffer
str.getChars(0, len - 1, buffer, 0);
// find decimal point
int point_idx = len;
for( int i = 0; i < len; i++ ) {
if( buffer[i] == '.' ) {
point_idx = i;
break;
}
}
// process all digits before the decimal point
float res = 0;
for( int i = 0; i < point_idx ; i++ ) {
int digit = buffer[i] - '0';
float value = digit * values[point_idx - i - 1];
res += value;
}
// process all digits after the decimal point
for( int i = point_idx + 1; i < len; i++ ) {
int digit = buffer[i] - '0';
float value = digit / values[i - point_idx];
res += value;
}
return res;
}
}
On Apr 14, 4:47 pm, Paul <[email protected]> wrote:
> OK, have stripped the FloatingPointParser bare, but can't run it as
> the native method, parseFltImpl() is throwing an
> UnsatisfiedLinkError...
>
> Maybe a silly question, but any ideas what to do from here?
>
> On Apr 14, 4:57 pm, Paul <[email protected]> wrote:
>
>
>
>
>
>
>
> > Sorry, should have mentioned this, this is 2.2 running on a Galaxy
> > Tab.
>
> > Based on Mark and Kostya's suggestions, I might just try to trim the
> > source and see what I can come up with - might be easier than managing
> > a mapping structure and will surely be lighter memory wise.
>
> > Will post back here with results as I have them!
>
> > Paul
>
> > On Apr 14, 4:26 pm, Dianne Hackborn <[email protected]> wrote:
>
> > > What version of the platform are you running on? It looks like in the
> > > current source most of the function is written in native code, but there
> > > is
> > > no native code in your profiling. I don't work on Dalvik, so I don't know
> > > when the native impl appeared.
>
> > > If this is the original Harmony implementation, it is probably way
> > > over-engineered and inefficient. :) If your numbers are normally of the
> > > form "xxxxx.yyy" you could try writing your own parser that parses that
> > > form
> > > as efficiently as possible, and bail to the generic impl if it finds
> > > something it can't handle.
>
> > > You certainly should be able to write something that doesn't have a
> > > $#&^%!!
> > > StringBuilder involved (good ghod!).
>
> > > On Thu, Apr 14, 2011 at 11:43 AM, Paul <[email protected]> wrote:
> > > > Ah, I see, create a kind of mapping from String -> float, and first
> > > > check the mapping before performing the conversion?
>
> > > > Here is what traceview is telling me for the calls to Float.valueOf():
>
> > > > java/lang/Float.valueOf (Ljava/lang/String;)Ljava/lang/Float;
> > > > * self = 1.9%
> > > > * java/lang/Float.parseFloat (Ljava/lang/String;)F = 94.3%
>
> > > > ** self = 1.3%
> > > > ** org/apache/harmony/luni/util/FloatingPointParser.parseFloat (Ljava/
> > > > lang/String;)F = 98.7%
>
> > > > *** self = 6.5%
> > > > *** org/a.../h.../luni/util/FloatingPointParser.initialParse (Ljava/
> > > > lang/String;I)Lorg/a.../h.../luni/util/FloatingPointParser
> > > > $StringExponentPair = 58.7%
>
> > > > **** self = 29.9%
> > > > **** a bunch of StringBuilder calls ~= 60%
>
> > > > *** java/lang/String.toLowerCase ()Ljava/lang/String = 26.8%
>
> > > > * java/lang/Float.valueOf (F)Ljava/lang/Float; = 3.8%
>
> > > > ** self = 48.4%
> > > > ** java/lang/Float.<init> (F)V = 51.2%
>
> > > > *** self = 69.7%
> > > > *** java/lang/Number.<init> ()V = 30.3%
>
> > > > Paul
>
> > > > On Apr 14, 2:20 pm, Dianne Hackborn <[email protected]> wrote:
> > > > > I think this may be what Marcin was suggesting -- if the same float
> > > > > value
> > > > > tends to appear multiple times, then when you parse a new one put it
> > > > > into
> > > > a
> > > > > cache. When you read strings, first see if the string is already in
> > > > > the
> > > > > cache to retrieve the previously parsed float value for it.
>
> > > > > Also can you get profiling data for where most of the time is spent
> > > > inside
> > > > > of Float.parseFloat()? In at least the current code base, the bulk of
> > > > the
> > > > > work for that is already done in native code. It would be
> > > > > interesting to
> > > > > know if most of your time is spent in that native function or the glue
> > > > > around it.
>
> > > > > It would also be useful to know what kind of CPU and what OS version
> > > > > you
> > > > are
> > > > > testing this on. For example, the JIT in newer platforms may
> > > > significantly
> > > > > improve performance here; CPUs that don't have hardware floating point
> > > > > support may have significantly worse performance.
>
> > > > > On Thu, Apr 14, 2011 at 9:17 AM, Paul <[email protected]> wrote:
> > > > > > The XML is the storage venue - Paths are drawn to the screen from
> > > > > > the
> > > > > > XML, and users can add new Paths (written back to the XML) or delete
> > > > > > existing Paths (data is deleted from XML).
>
> > > > > > On Apr 14, 12:11 pm, Marcin Orlowski <[email protected]>
> > > > > > wrote:
> > > > > > > > Any suggestions welcome!
>
> > > > > > > How often the source data changes? Do you reuse it later? If so
> > > > > > > and
> > > > > > source
> > > > > > > data changes not too often, cache it (write down processed data
> > > > > > > for
> > > > > > further
> > > > > > > reuse). If changes more frequently does it change totally or just
> > > > > > > is
> > > > > > small
> > > > > > > portion? If the latter, process only what's new. And cache it
> > > > > > > again.
>
> > > > > > > Regards,
> > > > > > > Marcin Orlowski
>
> > > > > > > *Tray Agenda <http://bit.ly/trayagenda>* - keep you daily schedule
> > > > > > handy...
> > > > > > > *Date In Tray* <http://bit.ly/dateintraypro> - current date at
> > > > glance...
> > > > > > > WebnetMobile on *Facebook <http://webnetmobile.com/fb/>* and
> > > > > > > *Twitter<http://webnetmobile.com/twitter/>
> > > > > > > *
>
> > > > > > --
> > > > > > You received this message because you are subscribed to the Google
> > > > > > Groups "Android Developers" group.
> > > > > > To post to this group, send email to
> > > > [email protected]
> > > > > > To unsubscribe from this group, send email to
> > > > > > [email protected]
> > > > > > For more options, visit this group at
> > > > > >http://groups.google.com/group/android-developers?hl=en
>
> > > > > --
> > > > > Dianne Hackborn
> > > > > Android framework engineer
> > > > > [email protected]
>
> > > > > Note: please don't send private questions to me, as I don't have time
> > > > > to
> > > > > provide private support, and so won't reply to such e-mails. All such
> > > > > questions should be posted on public forums, where I and others can
> > > > > see
> > > > and
> > > > > answer them.
>
> > > > --
> > > > You received this message because you are subscribed to the Google
> > > > Groups "Android Developers" group.
> > > > To post to this group, send email to [email protected]
> > > > To unsubscribe from this group, send email to
> > > > [email protected]
> > > > For more options, visit this group at
> > > >http://groups.google.com/group/android-developers?hl=en
>
> > > --
> > > Dianne Hackborn
> > > Android framework engineer
> > > [email protected]
>
> > > Note: please don't send private questions to me, as I don't have time to
> > > provide private support, and so won't reply to such e-mails. All such
> > > questions should be posted on public forums, where I and others can see
> > > and
> > > answer them.
--
You received this message because you are subscribed to the Google
Groups "Android Developers" group.
To post to this group, send email to [email protected]
To unsubscribe from this group, send email to
[email protected]
For more options, visit this group at
http://groups.google.com/group/android-developers?hl=en
