visualdatatools.com

[Adam]

Over the last few days, I've been working on a Python solution for writing .dtbin files. If anyone's interested in the work-in-progress, you can check it out with subversion:

[Adam]

I just realized that I can use this to implement modules, by following the conventions used for naming (Seq_Var for return variables). The great thing is that the .dtmod actually stores the script inside, so you can edit the source without recompiling or resaving the module . The creation process isn't nearly as slick as the magic Xcode project creation, though, so it may help to create a C++ project to get the variable names right.

I've posted a couple examples of how to use this to create modules (one of them works around problems linking against libDT.a on x86_64). The more basic one is http://code.google.com/p/mactlmgr/source/browse/datatank_py/examples/format_number.py, which converts a number to a string using printf-style formatting. These assume you've checked out the source to a directory searched by Python, or created a symlink so it's picked up.

[Adam]

In case anyone's following along, I just fixed (I think) a critical bug in saving 3D arrays. Row-major vs. column-major ordering in NumPy confuses me, so any advice or checks would be welcome.

I've also added a protocol for saving high-level DTSource objects such as DTMesh2D and DTStructuredMesh3D, and have slowly been adding Python implementations of those as I've needed them (only for saving, not for computations).

[Adam]

I've just landed a major bug fix in my DTDataFile Python code, so if you're using it, make sure to update. The short summary is that mixing reads and writes could cause file corruption under some conditions.

I noticed this while adding checks to make sure time variables are strictly increasing, as well as ensuring that your variable name time indices start at zero and are strictly increasing. Both of those have bitten me in DataTank, so now I can catch them earlier .

[Adam]

I've started adding read support for some high-level objects, using a class method. To use it, you'd do something like