Merge branch 'main' into bilback

Author: ColmTalbot

.github/workflows/latest-dynesty.yml

@@ -0,0 +1,55 @@
+# This workflow will install Python dependencies, run tests and lint with a variety of Python versions
+# For more information see: https://docs.github.com/en/actions/automating-builds-and-tests/building-and-testing-python
+
+name: Test against latest dynesty
+
+on:
+  push:
+    branches: [ "main" ]
+  pull_request:
+    branches: [ "main" ]
+  merge_group:
+  release:
+    types:
+      - published
+
+concurrency:
+  group: ${{ github.workflow }}-${{ github.ref }}
+  cancel-in-progress: true
+
+env:
+  CONDA_PATH: /opt/conda/
+
+jobs:
+  build:
+
+    name: ${{ matrix.python.name }} unit tests
+    runs-on: ubuntu-latest
+    container: ghcr.io/bilby-dev/bilby-python${{ matrix.python.short-version }}:latest
+    strategy:
+      fail-fast: false
+      matrix:
+        python:
+          - name: Python 3.12
+            version: 3.12
+            short-version: 312
+
+    steps:
+    - uses: actions/checkout@v4
+      with:
+        fetch-depth: 0
+        fetch-tags: true
+    - name: Install package
+      run: |
+        # activate env so that conda list shows the correct environment
+        source $CONDA_PATH/bin/activate python${{ matrix.python.short-version }}
+        python -m pip install .
+        python -m pip install git+https://github.com/joshspeagle/dynesty@master
+        conda list --show-channel-urls
+      shell: bash
+    - name: Run unit tests
+      run: |
+        python -m pytest --durations 10 -k dynesty
+    - name: Run sampler tests
+      run: |
+        python -m pytest test/integration/sampler_run_test.py --durations 10 -v -k dynesty

AUTHORS.md

@@ -1,8 +1,11 @@
 # Authors
 
 This file lists all the authors in first-name alphabetical order who have
-contributed (either by code contribution or indirectly). If your name is not
-listed here, please contact anyone on this list and raise your concern.
+contributed (either by code contribution or indirectly) to Bilby while development
+happened on git.ligo.org. On GitHub, all contributors are tracked at
+https://github.com/bilby-dev/bilby/graphs/contributors and so this is no longer
+actively maintained. If your name is not listed here and you contributed at
+https://git.ligo.org/lscsoft/bilby, please open an issue.
 
 Abhirup Ghosh
 Aditya Vijaykumar

CHANGELOG.md

@@ -5,9 +5,60 @@ The original MRs are only visible on the [LIGO GitLab repository](https://git.li
 
 ## [Unreleased]
 
-### Deprecated
+## [2.7.0]
+
+There are a few significant changes/additions in this release along with minor changes and removals.
+
+### Major changes
+
+* `Likelihood` instances can now be called as `likelihood.log_likelihood(parameters)`, see [here](https://bilby-dev.github.io/bilby/parameters.html) for more information.
+* Support the new API in [`dynesty=3`](https://github.com/joshspeagle/dynesty/releases/tag/v3.0.0)
+* Add a [new `WaveformGenerator`](https://bilby-dev.github.io/bilby/api/bilby.gw.waveform_generator.GWSignalWaveformGenerator.html) capable of using arbitrary waveform models implemented through the `gwsignal` waveform interface.
+
+### Additions
+* Added WeightedCategorical-Prior by @JasperMartins in https://github.com/bilby-dev/bilby/pull/893
+* ENH: Implement DiscreteValues prior by @unkaktus in https://github.com/bilby-dev/bilby/pull/947
+* ENH: Allow no parameters as state by @ColmTalbot in https://github.com/bilby-dev/bilby/pull/941
+* ENH: add support for new dynesty api by @ColmTalbot in https://github.com/bilby-dev/bilby/pull/950
+* FEAT: add gwsignal waveform generator by @ColmTalbot in https://github.com/bilby-dev/bilby/pull/877
+
+### Fixes
+* BUG: Fix sampling efficiency warning by @fgittins in https://github.com/bilby-dev/bilby/pull/953
+* TYPO: fix random call in example by @ColmTalbot in https://github.com/bilby-dev/bilby/pull/973
+* BUG: Fix matched filter SNR calculation in time domain injection by @ColmTalbot in https://github.com/bilby-dev/bilby/pull/957
+* TST: mark whitened strain tests as flaky by @mj-will in https://github.com/bilby-dev/bilby/pull/987
+* MAINT: update file extension logic by @mj-will in https://github.com/bilby-dev/bilby/pull/962
+* BUG: fix a bug with the dynesty v3 interface using dynesty native sampling by @ColmTalbot in https://github.com/bilby-dev/bilby/pull/994
+* BUG: fix a bug where some priors fail to read due to missing conversion functions by @asb5468 https://github.com/bilby-dev/bilby/pull/940
+* BUG: fix how kwargs are passed to custom dynesty samplers by @ColmTalbot https://github.com/bilby-dev/bilby/pull/999
+
+### Changes
+* ENH: plot_multiple: Allow plotting onto user-defined figure by @unkaktus in https://github.com/bilby-dev/bilby/pull/946
+* MAINT: change RNG imports by @mj-will in https://github.com/bilby-dev/bilby/pull/943
+* MAINT: np.trapz -> np.trapezoid by @ColmTalbot in https://github.com/bilby-dev/bilby/pull/974
+* MAINT: define Planck15-LAL cosmology using LAL constants by @mj-will in https://github.com/bilby-dev/bilby/pull/932
+* DEV: make sure all priors return float when needed by @ColmTalbot in https://github.com/bilby-dev/bilby/pull/979
+* Replace pytables with h5py by @duncanmmacleod in https://github.com/bilby-dev/bilby/pull/982
+
+### Deprecations
+* DEP: deprecate dnest4 interface by @ColmTalbot in https://github.com/bilby-dev/bilby/pull/980
+
+### Removed
+* MAINT: remove unsupported roq json weight file format by @ColmTalbot in https://github.com/bilby-dev/bilby/pull/945
+
+### Other changes
+* DOC: Fixed README link for opening bilby_pipe issues by @mick-wright in https://github.com/bilby-dev/bilby/pull/922
+* DOC: Correct docs for use_ratio argument of run_sampler by @mattpitkin in https://github.com/bilby-dev/bilby/pull/958
+* CI: add merge_group to enable merge queue by @mj-will in https://github.com/bilby-dev/bilby/pull/983
+* TST: remove dnest4 import test by @ColmTalbot in https://github.com/bilby-dev/bilby/pull/984
+* BLD: migrate to pyproject.toml by @mj-will in https://github.com/bilby-dev/bilby/pull/952
+* BLD: use release branches by @mj-will in https://github.com/bilby-dev/bilby/pull/954
+
+### New Contributors
+* @fgittins made their first contribution in https://github.com/bilby-dev/bilby/pull/953
+* @duncanmmacleod made their first contribution in https://github.com/bilby-dev/bilby/pull/982
 
-- `dnest4` is no longer tested, the plugin package should be used instead.
+**Full Changelog**: https://github.com/bilby-dev/bilby/compare/v2.6.0...v2.7.0
 
 ## [2.6.0]
 
@@ -1184,7 +1235,9 @@ First `pip` installable version https://pypi.org/project/BILBY/ .
 - All chainconsumer dependency as this was causing issues.
 
 
-[Unreleased]: https://github.com/bilby-dev/bilby/compare/v2.5.2...main
+[Unreleased]: https://github.com/bilby-dev/bilby/compare/v2.7.0...main
+[2.7.0]: https://github.com/bilby-dev/bilby/compare/v2.6.0...v2.7.0
+[2.6.0]: https://github.com/bilby-dev/bilby/compare/v2.5.2...v2.6.0
 [2.5.2]: https://github.com/bilby-dev/bilby/compare/v2.5.1...v2.5.2
 [2.5.1]: https://github.com/bilby-dev/bilby/compare/v2.5.0...v2.5.1
 [2.5.0]: https://github.com/bilby-dev/bilby/compare/v2.4.0...v2.5.0

bilby/core/prior/analytical.py

@@ -1412,33 +1412,73 @@ def cdf(self, val, *, xp=np):
         return xp.clip(result, 0, 1)
 
 
-class DiscreteValues(Prior):
-    def __init__(self, values, name=None, latex_label=None,
-                 unit=None, boundary="periodic"):
-        """ An equal-weighted discrete-valued prior
+class WeightedDiscreteValues(Prior):
+    def __init__(
+        self,
+        values,
+        weights=None,
+        name=None,
+        latex_label=None,
+        unit=None,
+        boundary="periodic",
+    ):
+        """A weighted discrete-valued prior
 
         Parameters
         ==========
-        values: array
+        values: 1d array_like of numeric type
             The discrete values of the prior.
+        weights: 1d array_like of numeric type or None
+            The weights of each category. If None, then all values are
+            equally weighted. Default: None.
         name: str
-            See superclass
+            The name of the parameter
         latex_label: str
-            See superclass
+            The latex label of the parameter. Used for plotting.
         unit: str
-            See superclass
+            The unit of the parameter. Used for plotting.
         """
+
         nvalues = len(values)
+        values = np.array(values)
+        if values.shape != (nvalues,):
+            raise ValueError(
+                f"Shape of argument 'values' must be 1d array-like but has shape {values.shape}"
+            )
         minimum = np.min(values)
         # Small delta added to help with MCMC walking
         maximum = np.max(values) * (1 + 1e-15)
-        super(DiscreteValues, self).__init__(
+        super(WeightedDiscreteValues, self).__init__(
             name=name, latex_label=latex_label, minimum=minimum,
             maximum=maximum, unit=unit, boundary=boundary)
         self.nvalues = nvalues
-        self.values = np.sort(np.array(values))
-        self.p = 1 / self.nvalues
-        self.lnp = -np.log(self.nvalues)
+        sorter = np.argsort(values)
+        self._values_array = values[sorter]
+
+        # inititialization of priors from repr only supports
+        # python buildins
+        self.values = self._values_array.tolist()
+
+        weights = (
+            np.array(weights) / np.sum(weights)
+            if weights is not None
+            else np.ones(self.nvalues) / self.nvalues
+        )
+        # check for consistent shape of input
+        if weights.shape != (self.nvalues,):
+            raise ValueError(
+                "Inconsistent shape of weights and number of values:"
+                + f"weights has shape {weights.shape} "
+                + f"while number of values is {self.values}"
+            )
+        self._weights_array = weights[sorter]
+        self.weights = self._weights_array.tolist()
+        self._lnweights_array = np.log(self._weights_array)
+
+        # save cdf for rescaling
+        _cumulative_weights_array = np.cumsum(self._weights_array)
+        # insert 0 for values smaller than minimum
+        self._cumulative_weights_array = np.insert(_cumulative_weights_array, 0, 0)
 
     def rescale(self, val):
         """
@@ -1455,8 +1495,22 @@ def rescale(self, val):
         =======
         Union[float, array_like]: Rescaled probability
         """
-        idx = np.asarray(np.floor(val * self.nvalues), dtype=int)
-        return self.values[idx]
+        index = np.searchsorted(self._cumulative_weights_array[1:], val)
+        return self._values_array[index]
+
+    def cdf(self, val):
+        """Return the cumulative prior probability of val.
+
+        Parameters
+        ==========
+        val: Union[float, int, array_like]
+
+        Returns
+        =======
+        float: cumulative prior probability of val
+        """
+        index = np.searchsorted(self._values_array, val, side="right")
+        return self._cumulative_weights_array[index]
 
     def prob(self, val):
         """Return the prior probability of val.
@@ -1469,17 +1523,11 @@ def prob(self, val):
         =======
         float: Prior probability of val
         """
-        if isinstance(val, (float, int)):
-            if val in self.values:
-                return self.p
-            else:
-                return 0
-        else:
-            val = np.atleast_1d(val)
-            probs = np.zeros_like(val, dtype=np.float64)
-            idxs = np.isin(val, self.values)
-            probs[idxs] = self.p
-            return probs
+        index = np.searchsorted(self._values_array, val)
+        index = np.clip(index, 0, self.nvalues - 1)
+        p = np.where(self._values_array[index] == val, self._weights_array[index], 0)
+        # turn 0d numpy array to scalar
+        return p[()]
 
     def ln_prob(self, val):
         """Return the logarithmic prior probability of val
@@ -1493,24 +1541,86 @@ def ln_prob(self, val):
         float:
 
         """
-        if isinstance(val, (float, int)):
-            if val in self.values:
-                return self.lnp
-            else:
-                return -np.inf
-        else:
-            val = np.atleast_1d(val)
-            probs = -np.inf * np.ones_like(val, dtype=np.float64)
-            idxs = np.isin(val, self.values)
-            probs[idxs] = self.lnp
-            return probs
+        index = np.searchsorted(self._values_array, val)
+        index = np.clip(index, 0, self.nvalues - 1)
+        lnp = np.where(
+            self._values_array[index] == val, self._lnweights_array[index], -np.inf
+        )
+        # turn 0d numpy array to scalar
+        return lnp[()]
 
 
-class Categorical(DiscreteValues):
-    def __init__(self, ncategories, name=None, latex_label=None,
+class DiscreteValues(WeightedDiscreteValues):
+    def __init__(self, values, name=None, latex_label=None,
                  unit=None, boundary="periodic"):
-        """ An equal-weighted Categorical prior
+        """An equal-weighted discrete-valued prior
 
+        Parameters
+        ==========
+        values: array
+            The discrete values of the prior.
+        name: str
+            See superclass
+        latex_label: str
+            See superclass
+        unit: str
+            See superclass
+        """
+        weights = np.ones_like(values)
+        super(DiscreteValues, self).__init__(
+            values=values,
+            weights=weights,
+            name=name,
+            latex_label=latex_label,
+            unit=unit,
+            boundary=boundary,
+        )
+
+
+class WeightedCategorical(WeightedDiscreteValues):
+    def __init__(
+        self,
+        ncategories,
+        weights=None,
+        name=None,
+        latex_label=None,
+        unit=None,
+        boundary="periodic",
+    ):
+        """A weighted Categorical prior
+
+        Parameters
+        ==========
+        ncategories: int
+            The number of available categories. The prior mass support is then
+            integers [0, ncategories - 1].
+        weights: 1d array_like or None
+            The weights of each category. If None, then all categories are
+            equally weighted. Default None.
+        name: str
+            The name of the parameter
+        latex_label: str
+            The latex label of the parameter. Used for plotting.
+        unit: str
+            The unit of the parameter. Used for plotting.
+        """
+        self.ncategories = ncategories
+        values = np.arange(0, ncategories)
+        super(WeightedCategorical, self).__init__(
+            values=values,
+            weights=weights,
+            name=name,
+            latex_label=latex_label,
+            unit=unit,
+            boundary=boundary,
+        )
+
+
+class Categorical(DiscreteValues):
+    def __init__(
+        self, ncategories, name=None, latex_label=None, unit=None, boundary="periodic"
+    ):
+        """An equal-weighted Categorical prior
         Parameters
         ==========
         ncategories: int
@@ -1523,9 +1633,15 @@ def __init__(self, ncategories, name=None, latex_label=None,
         unit: str
             See superclass
         """
+        self.ncategories = ncategories
         values = np.arange(0, ncategories)
-        DiscreteValues.__init__(self, values=values, name=name, latex_label=latex_label,
-                                unit=unit, boundary=boundary)
+        super(Categorical, self).__init__(
+            values=values,
+            name=name,
+            latex_label=latex_label,
+            unit=unit,
+            boundary=boundary,
+        )
 
 
 class Triangular(Prior):