[Matplotlib-users] rec2csv - forcing precision

[asayeed <as...@mb...>]

Hi there, 

It seems a little odd to me that rec2csv would overthrow float formatting,
such as FormatFloat(2).  I am reading in a file that looks like (with
csv2rec):

ES2006a C       2_1     if      23.53   23.66   0.13    0.17    -0.7844
-2.2863 IN      4.819   1.981   2.838   if      0       0       
ES2006a C       2_1     it      23.66   23.74   0.08    0.21    -1.3039
-1.7667 PRP     2.802   1.518   1.284   it      0       0       
ES2006a C       2_1     was     23.74   23.9    0.16    0.27    -0.663 
-2.5537 AUX     2.694   0.955   1.739   was     0       0       
ES2006a C       2_1     sent    23.9    24.17   0.27    0.36    -2.62  
-4.062  VBN     6.964   1.317   5.647   sent    1       2       

And I am doing a few numpy things to it, and then spitting the results back
out into a file that ends up looking like this:

ES2006a C       2_1     if      23.530000000000001      23.66   0.13   
0.170000
00000000001     -0.78439999999999999    -2.2863000000000002     IN     
4.819   
1.9810000000000001      2.8380000000000001      if      0       0       0.0
ES2006a C       2_1     it      23.66   23.739999999999998     
0.08000000000000
0002    0.20999999999999999     -1.3039000000000001     -1.7666999999999999     
PRP     2.802   1.518   1.284   it      0       0       0.0
ES2006a C       2_1     was     23.739999999999998      23.899999999999999      
0.16    0.27000000000000002     -0.66300000000000003    -2.5537000000000001    
AUX     2.694   0.95499999999999996     1.7390000000000001      was     0      
0       0.0
ES2006a C       2_1     sent    23.899999999999999      24.170000000000002     
0.27000000000000002     0.35999999999999999     -2.6200000000000001    
-4.0620000000000003     VBN     6.964   1.3169999999999999     
5.6470000000000002      sent    1       2       0.0

which is not at all what I want.  I want the numbers to look like the
originally did in the input file...or at least better truncated.

I tried to use FormatFloat(n), but the documentation (plus the mlib code)
says that rec2csv overrides the FormatFloat to get a %r format.  

My stopgap measure is to create my own FormatFloatForce that looks like
this:

class FormatFloatForce(mlab.FormatFormatStr):
    def __init__(self):
        mlab.FormatFormatStr.__init__(self, "%g")

    def toval(self, x):
	return x

    def fromstr(self, s):
        return float(s)

But this seems rather hideous to me, especially since I have to apply it to
every numerical column.  Am I simply doing this all wrong?

Yours,
--Asad.



--
View this message in context: http://matplotlib.1069221.n5.nabble.com/rec2csv-forcing-precision-tp40655.html
Sent from the matplotlib - users mailing list archive at Nabble.com.