Merge branch 'develop' into documentation

CHANGELOG.md

@@ -2,30 +2,39 @@
 
 All notable changes to this project will be documented in this file.
 
-## [2.0.0] - 2021-??-??
+## [2.0.0] - 2022-??-??
 ### Added
+- The possibility to work with Monte Carlo histories which are evenly or unevenly spaced was added.
+- `cov_Obs` added as a possibility to propagate the error of non Monte Carlo data together with Monte Carlo data.
 - `CObs` class added which can handle complex valued Markov chain Monte Carlo data and the corresponding error propagation.
 - Matrix to matrix operations like the matrix inverse now also work for complex matrices and matrices containing entries that are not `Obs` but `float` or `int`.
-- The possibility to work with Monte Carlo histories which are evenly or unevenly spaced was added.
+- Support for a new `json.gz` file format was added
 - The Corr class now has additional methods like `reverse`, `T_symmetry`, `correlate` and `reweight`.
-- `linalg` module now has explicit functions `inv` and `cholesky`.
+- `Corr.m_eff` can now cope with periodic and anti-periodic correlation functions
+- Forward, backward and improved variants of the first and second derivative were added to the `Corr` class
+- The `linalg` module now has explicit functions `inv` and `cholesky`.
 - `Obs` objects now have methods `is_zero` and `is_zero_within_error` as well as overloaded comparison operations.
 - Functions to convert Obs data to or from jackknife was added.
 - Alternative matrix multiplication routine `jack_matmul` was added to `linalg` module which makes use of the jackknife approximation and is much faster for large matrices.
 - Additional input routines for npr data added to `input.hadrons`.
+- The `sfcf` and `openQCD` input modules can now handle all recent file type versions.
+- `extract_t0` can now visualize the extraction on the fly
+- Module added which provides the Dirac gamma matrices in the Grid convention.
 - Version number added.
 
 ### Changed
 - The internal bookkeeping system for ensembles/replica was changed. The separator for replica is now `|`.
 - The fit functions were renamed to `least_squares` and `total_least_squares`.
+- The output of the fit functions is now a dedicated results class which keeps track of all relevant information
 - The fit functions can now deal with provided covariance matrices.
 - The convention for the fit range in the Corr class has been changed.
-- Obs.print was renamed to Obs.details and the output was improved.
+- Various method of the `Corr` class were renamed
+- `Obs.print` was renamed to `Obs.details` and the output was improved.
 - The default value for `Corr.prange` is now `None`.
 - The `input` module was restructured to contain one submodule per data source.
 - Performance of Obs.__init__ improved.
 
-### Deprecated
+### Removed
 - The function `plot_corrs` was deprecated as all its functionality is now contained within `Corr.show`
 - The kwarg `bias_correction` in `derived_observable` was removed
 - Obs no longer have an attribute `e_Q`

pyerrors/correlators.py

@@ -376,6 +376,24 @@ class Corr:
         """Reverse the time ordering of the Corr"""
         return Corr(self.content[:: -1])
 
+    def thin(self, spacing=2, offset=0):
+        """Thin out a correlator to suppress correlations
+
+        Parameters
+        ----------
+        spacing : int
+            Keep only every 'spacing'th entry of the correlator
+        offset : int
+            Offset the equal spacing
+        """
+        new_content = []
+        for t in range(self.T):
+            if (offset + t) % spacing != 0:
+                new_content.append(None)
+            else:
+                new_content.append(self.content[t])
+        return Corr(new_content)
+
     def correlate(self, partner):
         """Correlate the correlator with another correlator or Obs
 

tests/correlators_test.py

@@ -283,3 +283,11 @@ def test_hankel():
     corr.Hankel(2)
     corr.Hankel(6, periodic=True)
 
+
+def test_thin():
+    c = pe.Corr([pe.pseudo_Obs(i, .1, 'test') for i in range(10)])
+    c *= pe.cov_Obs(1., .1, '#ren')
+    thin = c.thin()
+    thin.fit(lambda a, x: a[0] * x)
+    c.thin(offset=1)
+    c.thin(3, offset=1)