Merge pull request #9 from hpc4cmb/multi_channel

Add support for FLAC compression of 64bit integers.

Author: tskisner

.github/workflows/test.yml

@@ -93,17 +93,28 @@ jobs:
             && conda activate test \
             && pip install .
 
-      - name: Run Serial Tests
+      - name: Serial Tests (MPI Disabled)
         run: |
           source ~/conda/etc/profile.d/conda.sh \
             && conda activate test \
             && export OMP_NUM_THREADS=2 \
+            && export MPI_DISABLE=1 \
             && mkdir -p test \
             && pushd test >/dev/null 2>&1 \
             && python3 -c 'import flacarray.tests; flacarray.tests.run()' \
             && popd >/dev/null 2>&1
 
-      - name: Run MPI Tests
+      - name: Serial Tests
+        run: |
+          source ~/conda/etc/profile.d/conda.sh \
+            && conda activate test \
+            && export OMP_NUM_THREADS=2 \
+            && mkdir -p test \
+            && pushd test >/dev/null 2>&1 \
+            && python3 -c 'import flacarray.tests; flacarray.tests.run()' \
+            && popd >/dev/null 2>&1
+
+      - name: MPI Tests
         run: |
           source ~/conda/etc/profile.d/conda.sh \
             && conda activate test \

docs/docs/cookbook.ipynb

@@ -10,12 +10,11 @@ timestreams that do not compress well with DEFLATE algorithms used by zip /
 gzip. This type of data is found in audio signals, scientific timestreams, etc.
 
 In the `flacarray` package we use only a small subset of features found in the
-libFLAC library. In particular, each data stream is compressed as a single, 32
-bit "channel". Stream data consisting of 32 bit integers (or 64 bit integers
-spanning a peak-to-peak range that fits into 32 bits) are compressed in a
-loss-less fashion. Floating point data is converted to 32 bit integers with a
-user-specified precision or quantization.
+libFLAC library. In particular, each data stream is compressed as either one or
+two 32 bit "channels". Stream data consisting of 32 or 64 bit integers are
+compressed in a loss-less fashion. Floating point data is converted to either 32
+or 64 bit integers with a user-specified precision or quantization.
 
 If you are specifically working with audio data and want to write flac format
-files, you should look at other software tools such as
+audio files, you should look at other software tools such as
 [pyflac](https://github.com/sonos/pyFLAC).

docs/docs/index.md

@@ -38,7 +38,7 @@ development libraries installed.
 If you have conda available, you can create an environment will all the
 dependencies you need to build flacarray from source. For this example, we
 create an environment called "flacarray". First create the env with all
-dependencies and activate it (FIXME, add a requirements file for dev):
+dependencies and activate it (FIXME: add a requirements file for dev):
 
     conda create -n flacarray \
         c-compiler numpy libflac cython meson-python pkgconfig
@@ -53,7 +53,7 @@ To build and install the package.
 
 To also work on docs, install additional packages:
 
-    conda install mkdocs mkdocstrings mkdocs-jupyter
+    conda install mkdocs mkdocstrings mkdocstrings-python mkdocs-jupyter
     pip install mkdocs-print-site-plugin
 
 ### Other Ways of Building