# VTKHDF SplitByComposite

**URL:** https://discourse.paraview.org/t/vtkhdf-splitbycomposite/17633
**Category:** ParaView Support
**Created:** [June 23, 2026, 12:21pm UTC](https://discourse.paraview.org/t/vtkhdf-splitbycomposite/17633 "2026-06-23T12:21:34Z")
**Posts on this page:** 5
**Page:** 1

<div class="post-metadata">

### Author: ![jfavre](https://discourse.paraview.org/letter_avatar_proxy/v4/letter/j/b5ac83/32.png) [@jfavre](https://discourse.paraview.org/u/jfavre)
#### Post date: [June 23, 2026, 12:21pm UTC](https://discourse.paraview.org/t/vtkhdf-splitbycomposite/17633/1 "2026-06-23T12:21:34Z")

</div>

I have multiple vtkUnstructuredGrids which have been grouped with GroupDatasets()

Attempting to save them in VTKHDF format, I use

SaveData(‘/foo.vtkhdf’, proxy=groupDatasets1, WriteAllTimeSteps=1,  
SplitByComposite=1)

The help button says that SplitByComposite would write each part of a vtkMultiBlockDataSet “_in their own file_”, such that each can be read individually; or did I misunderstand?

This is not happening for me. I see a **single** monolithic HDF5 file being written. I use v6.1.1

---

<div class="post-metadata">

### Author: ![mwestphal](https://discourse.paraview.org/user_avatar/discourse.paraview.org/mwestphal/32/17_2.png) [@mwestphal](https://discourse.paraview.org/u/mwestphal)
#### Post date: [June 23, 2026, 3:00pm UTC](https://discourse.paraview.org/t/vtkhdf-splitbycomposite/17633/2 "2026-06-23T15:00:22Z")

</div>

I just tested using ParaView 6.1.1, ParaView wrote three files as expected:

`mbsplit_Block_1.vtkhdf mbsplit_Block_2.vtkhdf mbsplit.vtkhdf `

The \_1 and \_2 files are valid .vtkhdf files.

---

<div class="post-metadata">

### Author: ![jfavre](https://discourse.paraview.org/letter_avatar_proxy/v4/letter/j/b5ac83/32.png) [@jfavre](https://discourse.paraview.org/u/jfavre)
#### Post date: [June 23, 2026, 4:28pm UTC](https://discourse.paraview.org/t/vtkhdf-splitbycomposite/17633/3 "2026-06-23T16:28:01Z")

</div>

I use the WriteAllTimeSteps=1 option. My data has Time. The attached Python code **does not** write three independent files.

[pvTestVTKHDFWriter.py](https://discourse.paraview.org/uploads/short-url/6TUVZNyM2GrrgX4sqDYOAWWcPQ8.py) (5.1 KB)

---

<div class="post-metadata">

### Author: ![mwestphal](https://discourse.paraview.org/user_avatar/discourse.paraview.org/mwestphal/32/17_2.png) [@mwestphal](https://discourse.paraview.org/u/mwestphal)
#### Post date: [June 24, 2026, 6:35am UTC](https://discourse.paraview.org/t/vtkhdf-splitbycomposite/17633/4 "2026-06-24T06:35:07Z")

</div>

@Louis_Gombert

---

<div class="post-metadata">

### Author: ![Louis\_Gombert](https://discourse.paraview.org/user_avatar/discourse.paraview.org/louis_gombert/32/14576_2.png) [@Louis\_Gombert](https://discourse.paraview.org/u/Louis_Gombert)
#### Post date: [June 24, 2026, 8:06am UTC](https://discourse.paraview.org/t/vtkhdf-splitbycomposite/17633/5 "2026-06-24T08:06:10Z")

</div>

SplitComposite/Partitions is not compatible with temporal datasets for now. I’ll make it clear in the documentation, all combinations of Split don’t have full compatibility with each other.