v0.21 Changelog

0.21

Released: Feb 8, 2022

[general] [feature] ¶
For ResStock the OpenStudio version has changed to v3.2.1. Also, the residential workflow generator has changed slightly. Simulation output files retention and deletion can be controlled through arguments to the ServerDirectoryCleanup measure.

References: pull request 240
[general] [feature] [eagle] ¶
The buildstock.csv is trimmed for each batch job to hold only the rows corresponding to buildings in the batch. This improves speed and memory consumption when the file is loaded in ResStock.

References: pull request 246
[general] [feature] ¶
For ResStock the OpenStudio version has changed to v3.3.0.

References: pull request 251
[general] [postprocessing] ¶
The output partition size of 4GB was making downstream data processing difficult. Both spark and dask clusters were failing due to out of memory errors. I’m changing it back to 1GB, which will make more files, but each will be more manageable.

References: pull request 247

[bugfix] ¶
There was a few days there when the version of some sublibrary (click) of dask was incompatible with the latest version of dask. We temporarily pinned the sublibrary so that new installs would work. They have fixed that problem now, so this removes the restriction on that library.

References: pull request 232
[bugfix] ¶
For ResStock the build_existing_model.sample_weight was inverse to what we would expect. The bug was identified in the residential workflow generator.

References: pull request 234
[bugfix] ¶
Fixes an issue that caused out of memory error when postprocessing large run with many upgrades.

References: #253, pull request 258, pull request 262
[bugfix] ¶
Fixes a bug that caused postprocessing to crash when there is only one datapoint.

References: #261, pull request 263
[bugfix] ¶
Fixes a bug that caused postprocessing to crash on small runs where some jobs have failed sims.

References: #265, pull request 266