FEATURE REQUEST: Install without saving/caching functionality #400
Labels
Comments
|
Dropping HDF5 dependency would not just mean that there is no cache, but also no data intake from HDF5 files, and, of course, no HDF5 output. This might be OK when we provide sufficient other means of reading data, e.g. from plain .npy or .csv |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Feature Description
While caching can be disabled, installing Acoular still depends on the packages that make caching possible and, to a greater extent, depends on packages that enable the saving and loading of large, static data such as
h5pyorpytables. This is not only bloat when a user explicitly does not want caching capabilities (see below for an example), but it also straight up makes installing Acoular impossible on some weaker/alternative platforms. For example, installing Acoular on the Google Coral edge TPU fails because ah5pydistribution for it does not exist and the device is too weak to compile it from source (it runs out of RAM). This is frustrating because the device is powerful enough to utilize Acoular's signal processing capabilities as-is with small in-memory data chunks, which renders any caching/saving methods unused and therefore not needed. Is it possible to install Acoular with an extra that omits these packages? For examplepip install acoular[no-cache]orpip install acoular[just-dsp]?Use Case
While "real-time" and "Python" may exclude each other, it is still very convenient to perform successive calculations on Raspberry Pi-esque single board computers in small memory chunks. There, I do not intend to analyze big data sets but rather one small buffer after the other for quasi real-time applications such as visualizations like Spectacoular already does, but on a smaller system.
Importance
Medium: Would be beneficial for some users.
The text was updated successfully, but these errors were encountered: