[NeXus-tech] [Fwd: Re: dimensions]

[Peterson, Peter F. (pf9)]

Pete,

There are a couple of things that will make this easier. Leaving off the
indices that aren't being linked/specified (what you have as a ':' in the
example) would be clearer (I think). Otherwise we should come up with a
different tag/attribute that makes the meaning of the ':' clearer. The other
thing that will make dimensions clearer/cleaner is to get inheritance like
what Mark has suggested so the rank of the data is fixed. TOFRAW is hideous
because of these two issues. Are we nearing a stage where inheritance makes
sense.

P^2


On 2/19/09 2:53 PM, "Pete Jemian" <prjemian at gmail.com> wrote:

> 
> perhaps:
> ---------------------------------% clip here
> %---------------------------------
>     <field name="final_wavelength" type="NX_FLOAT" units="NX_LENGTH">
>         <doc>Wavelength on leaving beamline component</doc>
>         <dimensions size="1">
>             <dim index="1">:</dim>
>         </dimensions>
>     </field>
>     <field name="final_beam_divergence" type="NX_FLOAT" units="NX_ANGLE">
>         <doc>Divergence FWHM of beam leaving this component</doc>
>         <dimensions size="2"><!-- [2,:] -->
>             <dim index="1">2</dim>
>             <dim index="2">:</dim>
>         </dimensions>
>     </field>
>     <field name="flux" type="NX_FLOAT" units="NX_FLUX">
>         <doc>flux incident on beam plane area</doc>
>         <dimensions size="1">
>             <dim index="1">i</dim>
>         </dimensions>
>     </field>
> ---------------------------------% clip here
> %---------------------------------
> 
> 
> 
> Pete Jemian wrote:
>> Regarding the definition of dimensions, so far, so good.
>> Now, here is part of (revised) NXbeam.nxdl.xml:
>> 
>> ---------------------------------% clip here
>> %---------------------------------
>>    <field name="final_wavelength_spread" type="NX_FLOAT" units="NX_LENGTH">
>>        <doc>Wavelength spread FWHM of beam leaving this component</doc>
>>        <dimensions size="[:]" />
>>    </field>
>>    <field name="final_beam_divergence" type="NX_FLOAT" units="NX_ANGLE">
>>        <doc>Divergence FWHM of beam leaving this component</doc>
>>        <dimensions size="2"><!-- [2,:] --></dimensions>
>>    </field>
>>    <field name="flux" type="NX_FLOAT" units="NX_FLUX">
>>        <doc>flux incident on beam plane area</doc>
>>        <dimensions size="i" />
>>    </field>
>> ---------------------------------% clip here
>> %---------------------------------
>> 
>> How do we handle this?  (So far, it is valid by the changes in NXDL rules.)
>> Also, we need better documentation strings than given below.
>> 
>> 
>> (Here is just the revised part of nxdl.xsd):
>> ---------------------------------% clip here
>> %---------------------------------
>>    <xs:complexType name="dimsType">
>>        <!--
>>            Here is a model:
>>            <field name="data">
>>                <dimensions size="3"/>
>>                ...
>>            </field><field name="time-of-flight">
>>                <dimensions size="1">
>>                    <dim index="1" ref="data" refindex="3" incr="1"/>
>>                </dimensions>
>>                ...
>>            </field>
>>           
>>            In old terms this would make data[i,j,k] and time-of-flight[k+1].
>>            We need to get to the stage where people
>>            don't need to read a manual to understand an existing NXDL.
>>        -->
>>        <xs:annotation>
>>            <xs:documentation>dimensions of a data element in a NeXus
>> file</xs:documentation>
>>        </xs:annotation>
>>        <xs:sequence>
>>            <xs:element name="dim" minOccurs="0" maxOccurs="unbounded">
>>                <xs:complexType>
>>                    <xs:attribute name="index" type="nx:NX_CHAR">
>>                        <xs:annotation>
>>                            <xs:documentation>Number indicating which
>> subscript is being referenced,
>>                                ranging from 1 up to RANK (rank of the data
>> structure).  For example,
>>                                given an array A[i,j,k], index="1" would refer
>> to the "i" subscript.</xs:documentation>
>>                        </xs:annotation>
>>                    </xs:attribute>
>>                    <xs:attribute name="ref" type="nx:NX_CHAR">
>>                        <xs:annotation>
>>                            <xs:documentation>Name of the data structure being
>> referenced 
>>                                   indirectly.</xs:documentation>
>>                        </xs:annotation>
>>                    </xs:attribute>
>>                    <xs:attribute name="refindex" type="nx:NX_CHAR">
>>                        <xs:annotation>
>>                            <xs:documentation>Index within the data structure
>> being 
>>                                  referenced indirectly.</xs:documentation>
>>                        </xs:annotation>
>>                    </xs:attribute>
>>                    <xs:attribute name="incr" type="nx:NX_CHAR">
>>                        <xs:annotation>
>>                            <xs:documentation>What is this?</xs:documentation>
>>                        </xs:annotation>
>>                    </xs:attribute>
>>                </xs:complexType>
>>            </xs:element>
>>        </xs:sequence>
>>        <xs:attribute name="size" type="nx:NX_CHAR">
>>            <xs:annotation>
>>                <xs:documentation>Rank (number of dimensions) of the data
>> structure. 
>>                    For example: a[5] has size="1" while b[8,5,6,4] has
>> size="4". 
>>                    See
>> http://en.wikipedia.org/wiki/Rank_(computer_programming)
>>                    for more details.</xs:documentation>
>>            </xs:annotation>
>>        </xs:attribute>
>>    </xs:complexType>
>> ---------------------------------% clip here
>> %---------------------------------
>> 
>> Pete
>> 
>> 
>> 
>> 
>>