Merge pull request #3173 from zm711/load-if-exists-docs

Add explainer of compute always computing in the analyzer (vs WaveformExtractor behavior)

Author: alejoe91

doc/modules/core.rst

@@ -360,6 +360,15 @@ examples is seen below:
         ["random_spikes", "waveforms", "templates", "noise_levels"]
     )
 
+Note that any extension entered into the :code:`compute` function will be computed. Thus, even if an extension has already been computed
+it will be recomputed when :code:`compute` is called even if the parameters are the same.
+To check if an extension has already been computed and reload it, you can simply do:
+
+.. code-block:: python
+
+    if sorting_analyzer.has_extension("templates"):
+        templates_extension = sorting_analyzer.get_extension("templates")
+
 It is important when calculating extensions to remember which backend you are using. :code:`compute` accepts an argument
 :code:`save` which will write results to disk if using the :code:`zarr` or :code:`binary_folder` backends. If your :code:`SortingAnalyzer`
 is in memory using :code:`save=True` **will not** write to disk since spikeinterface does not know where to save it.

examples/tutorials/core/plot_4_sorting_analyzer.py

@@ -129,7 +129,8 @@
 
 ###############################################################################
 # Note that there are a few different ways to calculate extensions. You might find it easier
-# to keep the calculation parameters in a dictionary. Let's recalculate everything using this notation:
+# to keep the calculation parameters in a dictionary. Let's recalculate everything using this notation
+# (note that extensions will be recomputed when we call compute even if the parameters are the same):
 
 job_kwargs = dict(n_jobs=8, chunk_duration="1s", progress_bar=True)
 compute_dict = {

examples/tutorials/waveform_extractor_to_sorting_analyzer.rst

@@ -349,6 +349,9 @@ looks slightly different. Let's calculate these extensions, and also add a param
             )
 
 Note that if a parent is recomputed, all its children are deleted to maintain data consistency.
+Also note that the ``load_if_exists`` semantic that was found in the ``WaveformExtractor`` has
+been removed from the ``SortingAnalyzer``. This means that each call to ``compute`` will recompute
+and overwrite the previous results. Adjust your scripts according to account for this new behavior.
 Read more about this, the extensions and their keyword arguments in the
 `postprocessing module documentation <https://spikeinterface.readthedocs.io/en/latest/modules/postprocessing.html>`_