[emboss-dev] computed maximum forbidden in ACD ?

Peter Rice pmr at ebi.ac.uk
Tue Aug 17 16:22:58 EDT 2010


Dear Guy,

> When I run blast I get :
>
> Error: File /OPT/emboss63/share/EMBOSS/acd/blast.acd line 100:
> (wordsize) Attribute failrange: required with any calculated min/max
>
> I am as good as certain that this behaviour has appeared with EMBOSS
> version 6.3.0. In the past it was allowed to set a "maximum" that
> depended on the choice of another parameter, and we can see that it
> could occasionally make sense, but this seems from now on forbidden. I
> this a bug or a feature ?

It is a fix for a feature. With calculated maximum or minimum values 
(e.g. depending on a window size) it was possible for the maximum to be 
less than the minimum. In such cases we could logically use either the 
maximum or the minimum - and some applications were found to require one 
choice, others needed the other.

After some discussion we decided to add extra attributes to control the 
behaviour. You can add two new attributes:

trueminimum: "N"  (if max/min overlap, use minimum}
failrange:   "Y"  (Fail if (calculated) ranges overlap}
rangemessage: ""   (Failure message if (calculated ranges) overlap}

A common solution (good for your case) is:

failrange: "N"
trueminimum: "Y"

By adding the error messages we made sure that an ACD file with a 
calculated range will give messages to the developer suggesting missing 
attributes to be added.

If you set failrange: "Y" you need to define a message explaining to the 
end user why the range might fail

If you set failrange: "N" the calculated range is accepted, but you also 
need to set trueminimum to say whether you want the minimum value to 
apply (usual to avoid getting negative values) or the maximum to avoid 
values going too large.

So, you get the "failrange is required" message. When you set that you 
get another message (depending whether it is true or false) telling you 
to set one of the other attributes as well.

Hope this makes it clearer!

Peter




More information about the emboss-dev mailing list