[Bioperl-l] no leading digit in expect values < 1e-99

Saurabh D. Patel patels02@doc.mssm.edu
Tue, 12 Jun 2001 23:25:56 -0400


I sent an email to the blast helpdesk and this is the reply I recieved:

>Hi;
>
>This is actually a feature and not a bug.  The following are the rules
>for formatting the expect value and your example falls into situation
>number 2
>
>1.) if the expect value is smaller than 1.0e-180 it is set to zero,
>
>2.) if the expect value is smaller than 1.0e-99 only the exponent is
>shown
>
>3.) if the expect value is smaller than 0.0009 it is formatted with
>with one significant digit (e.g., 2e-15)
>
>4.) if it is less than 0.1, it is formatted with three digits behind
>.the decimal point (0.075).
>
>5.) if it is less than 1.0 it is formatted with two digits behind the
>decimal point (0.56).
>
>6.) if it is less than 10 there is one digit behind the decimal point
>(9.6).
>
>These rules are a reflection of the accuracy of the results in view of
>the differences between the random sequence model that was used and the
>behavior of actual biological sequences.
>
>Sincerely, 
>
>Scott McGinnis, M.S.
>National Center of Biotechnology Information. 
>HTTP://www.ncbi.nlm.nih.gov


> Omitting the leading digit was a very very ill decision by NCBI IMO.
> You can't even parse e-108 in C sscanf("e-108","%lf",&double_var).
> Java (JDK 1.3) Double("e-108") doesn't do the job either. Who at NCBI
> was responsible for this decision?? Is there a newer revision
> correcting this?
> 
>         Hilmar
> 
> --
> -------------------------------------------------------------
> Hilmar Lapp                            email: lapp@gnf.org
> GNF, San Diego, Ca. 92121              phone: +1-858-812-1757
> -------------------------------------------------------------

So since there is no revision that corrects this feature, I think the
BPlite parser should take care of it.

Saurabh.