platform: intel macintosh, python2.5 , latest yaml.
Description:
depending on a floats value dump() will output a different representation format. Sometimes it uses a tag format and others a standard float notation depending on the value.
example:
In [160]:dat ="""
float1: 1.0E-8
float2: 1.0E-9
"""
In [165]:t = yaml.load(dat)
In [166]:t
Out[166]:{'float1': 1e-08, 'float2': 1.0000000000000001e-09}
In [167]:yaml.dump(t)
Out[167]:"{float1: !!float '1e-08', float2: 1.0000000000000001e-09}\n"
notice the !!float tag followed by a string.
regression:
this appears to be an attempt to compensate for fact that python's string representation of some float values having different formats:
that is in python2.5 the repr() of 1.0e-8 is actually 1e-08 (it supresses the decimal point). Since YAML does not recognize floats without the decimal point the dump command adds the float tag. However while the repr() of 1.0E-9 is 1.0000000000000001e-09. which has the decimal point making it legal YAML and it dump avoids the float tag
It seems to me that a more consistent and desirable result would be obtained if dump were to output 1e-8 as 1.0e-8 rather than using the float if one is parsing a yaml file in another language the tagged format of the float can't be read easily. (e.g. consider reading this in fortran or perl without using a yaml lib to read the file.)
