Cosmocnc support

pyproject.toml

@@ -42,6 +42,9 @@ mflike = [
     "fgspectra (>=1.3.0); python_version < '3.13'", # from mflike
     "syslibrary (>=0.2.0); python_version < '3.13'", # from mflike
 ]
+cosmocnc = [
+    "cosmocnc (>=1.0)",
+]
 docs = [
     "sphinx (<9); python_version < '3.12'", # tensorflow 2.13 pins typing-extensions==4.5.0
     "sphinx; python_version >= '3.12'",
@@ -55,6 +58,7 @@ dev = [
     "pre-commit",
 ]
 all = [
+    "soliket[cosmocnc]",
     "soliket[emulator]",
     "soliket[pyccl]",
     "soliket[pyhalomodel]",
@@ -128,3 +132,6 @@ exclude_lines = [
     # Don't complain about IPython completion helper
     "def _ipython_key_completions_",
 ]
+
+[tool.uv.sources]
+cosmocnc = { git = "https://github.com/ggalloni/cosmocnc.git", branch = "soliket_compatibility" }

tests/conftest.py

@@ -67,6 +67,16 @@ def check_skip_mflike():
     pytest.importorskip(modname="mflike", reason="Couldn't import 'mflike' module")
 
 
+@pytest.fixture
+def check_skip_cosmocnc():
+    """
+    Check if the cosmocnc module can be imported, otherwise skip the tests.
+    """
+    pytest.importorskip(
+        modname="cosmocnc", reason="Couldn't import 'cosmocnc' module"
+    )
+
+
 @pytest.fixture
 def install_planck_lite():
     """

tests/likelihood_refs.yaml

@@ -21,6 +21,27 @@ clusters:
   rtol: *rtol
   atol: *atol
 
+cosmocnc_clusters:
+  value: 99492.945
+  rtol: *rtol
+  atol: *atol
+cosmocnc_clusters_binned:
+  value: 49505.109
+  rtol: *rtol
+  atol: *atol
+cosmocnc_clusters_unbinned_backward:
+  value: 99544.550
+  rtol: *rtol
+  atol: *atol
+cosmocnc_clusters_multi_obs:
+  value: 76839.435
+  rtol: *rtol
+  atol: *atol
+cosmocnc_clusters_stacked_lensing:
+  value: 98985.873
+  rtol: *rtol
+  atol: *atol
+
 cosmopower:
   value: -295.139
   rtol: *rtol

tests/test_clusters.py

@@ -6,6 +6,140 @@
 import soliket.clusters.survey as survey
 from soliket.clusters import tinker
 
+
+def _get_cosmocnc_common_config():
+    """Common config shared across all cosmocnc cluster tests."""
+    import cosmocnc
+
+    base = cosmocnc.path_to_cosmocnc
+    survey_sr = os.path.join(base, "surveys", "survey_sr_so_sim.py")
+    survey_cat = os.path.join(base, "surveys", "survey_cat_so_sim.py")
+
+    cnc_common = {
+        "stop_at_error": True,
+        "survey_sr": survey_sr,
+        "survey_cat": survey_cat,
+        "cosmology_tool": "cobaya",
+        "power_spectrum_type": "cobaya",
+        "obs_select": "q_so_sim",
+        "cluster_catalogue": "SO_sim_0",
+        "n_points": 2048,
+        "n_z": 50,
+        "z_min": 0.01,
+        "z_max": 3.0,
+        "M_min": 5e13,
+        "M_max": 5e15,
+        "obs_select_min": 5.0,
+        "obs_select_max": 200.0,
+        "hmf_calc": "cnc",
+        "hmf_type": "Tinker08",
+        "mass_definition": "500c",
+        "hmf_type_deriv": "numerical",
+        "interp_tinker": "linear",
+        "cosmo_amplitude_parameter": "sigma_8",
+        "cosmo_param_density": "physical",
+        "cosmocnc_verbose": "none",
+        "number_cores_hmf": 1,
+        "number_cores_abundance": 1,
+        "number_cores_data": 1,
+        "number_cores_stacked": 1,
+    }
+    theory = {
+        "camb": {
+            "extra_args": {
+                "num_massive_neutrinos": 1,
+                "accurate_massive_neutrino_transfers": True,
+                "nonlinear": False,
+                "kmax": 50.0,
+            }
+        },
+    }
+    params = {
+        "ombh2": 0.0224,
+        "omch2": 0.12,
+        "H0": 67.0,
+        "logA": 3.05,
+        "As": {"value": "lambda logA: 1e-10*np.exp(logA)"},
+        "ns": 0.965,
+        "tau": 0.054,
+        "nnu": 3.046,
+        "mnu": 0.06,
+        "A_szifi": -4.3054,
+        "alpha_szifi": 1.12,
+        "sigma_lnq_szifi": 0.173,
+        "bias_sz": 0.8,
+        "dof": 0.0,
+    }
+    return cnc_common, theory, params
+
+
+def _get_cosmocnc_clusters_info(mode="unbinned"):
+    """CNCLike-based cluster likelihood setup for various modes."""
+    cnc_common, theory, params = _get_cosmocnc_common_config()
+
+    if mode == "unbinned":
+        cnc_like = dict(
+            **cnc_common,
+            observables=[["q_so_sim"]],
+            data_lik_from_abundance=True,
+            likelihood_type="unbinned",
+            stacked_likelihood=False,
+        )
+    elif mode == "binned":
+        cnc_like = dict(
+            **cnc_common,
+            observables=[["q_so_sim"]],
+            data_lik_from_abundance=True,
+            likelihood_type="binned",
+            stacked_likelihood=False,
+        )
+    elif mode == "unbinned_backward":
+        cnc_like = dict(
+            **cnc_common,
+            observables=[["q_so_sim"]],
+            data_lik_from_abundance=False,
+            likelihood_type="unbinned",
+            stacked_likelihood=False,
+        )
+    elif mode == "multi_obs":
+        cnc_like = dict(
+            **cnc_common,
+            observables=[["q_so_sim", "p_so_sim"]],
+            data_lik_from_abundance=True,
+            likelihood_type="unbinned",
+            stacked_likelihood=False,
+        )
+        params.update({
+            "bias_cmblens": 0.8,
+            "a_lens": 1.0,
+            "sigma_lnp": 0.2,
+            "corr_lnq_lnp": 0.0,
+        })
+    elif mode == "stacked_lensing":
+        cnc_like = dict(
+            **cnc_common,
+            observables=[["q_so_sim"]],
+            data_lik_from_abundance=False,
+            likelihood_type="unbinned",
+            stacked_likelihood=True,
+            stacked_data=["p_so_sim_stacked"],
+            compute_stacked_cov=True,
+        )
+        params.update({
+            "bias_cmblens": 0.8,
+            "a_lens": 1.0,
+            "sigma_lnp": 0.2,
+            "corr_lnq_lnp": 0.0,
+        })
+
+    return {
+        "likelihood": {"cosmocnc.CNCLike": cnc_like},
+        "theory": theory,
+        "params": params,
+        "sampler": {"evaluate": None},
+    }
+
+
 clusters_like_and_theory = {
     "likelihood": {"soliket.ClusterLikelihood": {"stop_at_error": True}},
     "theory": {
@@ -159,3 +293,51 @@ def test_dn_dlogM_small_grid():
     # Expect output shape (nM, nz)
     assert out.shape[0] == M.shape[0]
     assert out.shape[1] == z.shape[0]
+
+
+# --- cosmocnc-based cluster tests (no pyccl required) ---
+
+
+def test_cosmocnc_clusters_model(check_skip_cosmocnc):
+    info = _get_cosmocnc_clusters_info()
+    _ = get_model(info)
+
+
+def test_cosmocnc_clusters_loglike(check_skip_cosmocnc, likelihood_refs):
+    ref = likelihood_refs["cosmocnc_clusters"]
+    info = _get_cosmocnc_clusters_info()
+    model = get_model(info)
+    lnl = model.loglikes({})[0]
+    assert np.isclose(lnl, ref["value"], rtol=ref["rtol"], atol=ref["atol"])
+
+
+def test_cosmocnc_clusters_binned(check_skip_cosmocnc, likelihood_refs):
+    ref = likelihood_refs["cosmocnc_clusters_binned"]
+    info = _get_cosmocnc_clusters_info(mode="binned")
+    model = get_model(info)
+    lnl = model.loglikes({})[0]
+    assert np.isclose(lnl, ref["value"], rtol=ref["rtol"], atol=ref["atol"])
+
+
+def test_cosmocnc_clusters_unbinned_backward(check_skip_cosmocnc, likelihood_refs):
+    ref = likelihood_refs["cosmocnc_clusters_unbinned_backward"]
+    info = _get_cosmocnc_clusters_info(mode="unbinned_backward")
+    model = get_model(info)
+    lnl = model.loglikes({})[0]
+    assert np.isclose(lnl, ref["value"], rtol=ref["rtol"], atol=ref["atol"])
+
+
+def test_cosmocnc_clusters_multi_obs(check_skip_cosmocnc, likelihood_refs):
+    ref = likelihood_refs["cosmocnc_clusters_multi_obs"]
+    info = _get_cosmocnc_clusters_info(mode="multi_obs")
+    model = get_model(info)
+    lnl = model.loglikes({})[0]
+    assert np.isclose(lnl, ref["value"], rtol=ref["rtol"], atol=ref["atol"])
+
+
+def test_cosmocnc_clusters_stacked_lensing(check_skip_cosmocnc, likelihood_refs):
+    ref = likelihood_refs["cosmocnc_clusters_stacked_lensing"]
+    info = _get_cosmocnc_clusters_info(mode="stacked_lensing")
+    model = get_model(info)
+    lnl = model.loglikes({})[0]
+    assert np.isclose(lnl, ref["value"], rtol=ref["rtol"], atol=ref["atol"])