seissolxdm/seissolxdmfwriter: bugfixes

[Thomas-Ulrich]

• The extractor was not working with SR (computed from SRs and SRd), because the derived seissolxdmf was not passed to seissolxdmdwriter (and GetDataLocationPrecisionMemDimension needed to be specialized as well but was not). This required a change of the interface, so I increased version number of 0.4.0.
• seissolxdmf was not able to read a 1D array (e.g. partition, fault-tag) when the array was padded.
• seissolxdmfwriter was not working properly when specifiying a path in the prefix (in this case the basename should be written in the xdmf, to allow opening the file from any directory).
• dependency to tqdm is added.
• other bug fixes.

[davschneller]

Quick question: what does the 0 do here?

[Thomas-Ulrich]

Good question!
It turns out that there is a bug in the xdmfwriter of seissol (not the python module!), as partition or fault-tags are padded, but the size written in the xdmf file in the unpadded one.
Here is an example of output:

    </Geometry>
    <Time Value="0"/>
    <Attribute Name="partition" Center="Cell">
     <DataItem  NumberType="Int" Precision="4" Format="Binary" Dimensions="4368">tpv5-fault_cell/mesh0/partition.bin</DataItem>
    </Attribute>
    <Attribute Name="fault-tag" Center="Cell">
     <DataItem  NumberType="Int" Precision="4" Format="Binary" Dimensions="4368">tpv5-fault_cell/mesh0/fault-tag.bin</DataItem>
    </Attribute>
    <Attribute Name="SRs" Center="Cell">
     <DataItem ItemType="HyperSlab" Dimensions="4368">
      <DataItem NumberType="UInt" Precision="4" Format="XML" Dimensions="3 2">0 0 1 1 1 4368</DataItem>
      <DataItem NumberType="Float" Precision="4" Format="Binary" Dimensions="1 2097152">tpv5-fault_cell/mesh0/SRs.bin</DataItem>
     </DataItem>
    </Attribute>
    <Attribute Name="SRd" Center="Cell">
     <DataItem ItemType="HyperSlab" Dimensions="4368">
      <DataItem NumberType="UInt" Precision="4" Format="XML" Dimensions="3 2">0 0 1 1 1 4368</DataItem>
      <DataItem NumberType="Float" Precision="4" Format="Binary" Dimensions="1 2097152">tpv5-fault_cell/mesh0/SRd.bin</DataItem>
     </DataItem>
    </Attribute>

Therefore if we read with idt=-1, we read a much larger array than expected (and the inferred ndt is very large in the case above and myData cannot be reshaped).
As a work-around, I set idt=0 so that just the first chunk of size nElements get read.

[Thomas-Ulrich]

I've added a comment (but it would also make sense to fix the xdmfwriter)

[Thomas-Ulrich]

TUM-I5/XdmfWriter#13