Closed prjemian closed 2 years ago
Reduced times will benefit work on #65.
Some requirements will be moved to pip install.
Time to build (re)install conda environment
==================== ========== ==============================
phase elapsed, s comments
==================== ========== ==============================
Python core 29 no downloads
+testing and QA 33 no downloads
+Qt 35 no downloads
+general support 152 includes numpy 1.21.5 download
repeat 126 no downloads, still a long time
just 2 packages 65 bitshuffle, epics-base
-bitshuffle 36 only epics-base
+pyEpics 41
+pandas, pint 41
+ipympl, lxml 45
+pymongo, xlrd 45
+h5py 50
+imagecodecs-lite 101 <-- !
+scikit-image 89 time dropped
+HDF5 compression 129 bitshuffle, hdf5plugin
+Bluesky framework 866 <-- !
-hklpy, pydm 375 <-- !
-bluesky-widgets 338
+pydm 414 hklpy & bluesky-widgets take ~6m to solve (move to pip?)
+user requests 157 but no Bluesky framework packages
+packaging tools 248
+pip installs 178 but no packaging tools
+pip install 161 bluesky-widgets
+pip install 178 hklpy - pulls in more of Bluesky
+pip install 163 remaining Bluesky packages
final version 155
==================== ========== ==============================
More than 5x faster.
bash script to time each installation variation
#!/bin/bash
begin=$(date +%s)
conda env create --force -f ./entest.yml
end=$(date +%s)
echo "elapsed $(($end-$begin))"
Can the time to solve the conda environment requirements be reduced?