PanDDA 0.2.14: new module via ccp4-python pip, ccp4/9.0 dep

Installs panddas into its own prefix (--prefix + --ignore-installed) so the
shared ccp4/9.0 module is never modified; PYTHONPATH prepend shadows the
ccp4-bundled copy. ccp4/9.0 is build- and runtime dependency.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0152MuuYUaapP7CtxrDAfvBh

PanDDA/README.md

@@ -0,0 +1,22 @@
+# PanDDA
+
+Pan-Dataset Density Analysis (`panddas` from PyPI) for fragment-screening MX data.
+<https://pandda.bitbucket.io>
+
+Installed with ccp4/9.0's `ccp4-python -m pip`, but into this module's own prefix
+(`--prefix` + `--ignore-installed`) so the shared ccp4/9.0 install is never modified.
+The vendor's "uninstall bundled panddas / upgrade pip+numpy inside ccp4" steps are
+replaced by the modulefile's PYTHONPATH prepend, which shadows ccp4's bundled copy.
+ccp4/9.0 is both build- and runtime dependency (auto-loaded).
+
+Build on Ra: see repo README / new-module recipe (`modbuild build 0.2.14`).
+
+Verify after load:
+
+```bash
+ccp4-python -c 'import pandda; print(pandda.__file__)'   # must point into /opt/psi/MX/PanDDA
+pandda.analyse -h
+```
+
+Note: panddas 0.2.14 (2018) predates CCP4 9's Python 3.9; if it fails with py2
+syntax errors at runtime, the tool needs the CCP4 7-era python or PanDDA2 instead.

PanDDA/build

@@ -0,0 +1,36 @@
+#!/usr/bin/env modbuild
+
+pbuild::prep() {
+    :
+}
+
+pbuild::configure() {
+    :
+}
+
+pbuild::compile() {
+    :
+}
+
+pbuild::install() {
+    # ccp4/9.0 is a build_requires, so ccp4-python is on PATH here.
+    command -v ccp4-python >/dev/null || \
+        std::die 42 "PanDDA: ccp4-python not on PATH — ccp4/9.0 module did not load"
+
+    # Vendor doc says: uninstall bundled panddas, upgrade pip+numpy, pip install
+    # panddas — all of which would mutate the shared ccp4/9.0 module. Instead:
+    # --prefix keeps everything inside this module, --ignore-installed pulls
+    # fresh deps (incl. numpy) here rather than reusing/soiling ccp4's, and the
+    # modulefile's PYTHONPATH prepend shadows the ccp4-bundled panddas.
+    ccp4-python -m pip install --prefix="${PREFIX}" --ignore-installed \
+        "panddas==${V_PKG}" || \
+        std::die 42 "PanDDA: pip install failed — check network and ccp4-python version"
+
+    # stable alias so the modulefile needn't know ccp4's python X.Y
+    local -a sp=( "${PREFIX}"/lib/python*/site-packages )
+    [[ -d "${sp[0]}" ]] || std::die 42 "PanDDA: no site-packages under ${PREFIX}/lib"
+    ln -sfn "${sp[0]#"${PREFIX}"/}" "${PREFIX}/site-packages"
+
+    [[ -x "${PREFIX}/bin/pandda.analyse" ]] || \
+        std::die 42 "PanDDA: bin/pandda.analyse missing — install incomplete"
+}

PanDDA/files/config.yaml

@@ -0,0 +1,14 @@
+---
+format: 1
+PanDDA:
+  defaults:
+    group: MX
+    overlay: base
+    relstage: unstable
+    # no urls/shasums — pip fetches panddas==<version> from PyPI
+  versions:
+    0.2.14:
+      config:
+        relstage: unstable
+        build_requires: [ccp4/9.0]
+        runtime_deps: [ccp4/9.0]

PanDDA/modulefile

@@ -0,0 +1,16 @@
+#%Module1.0
+
+module-whatis       "PanDDA: multi-dataset crystallographic analysis for fragment screening"
+module-url          "https://pandda.bitbucket.io"
+module-license      "CC BY-SA 4.0"
+module-maintainer   "Jiaxin Duan <jiaxin.duan@psi.ch>"
+
+module-help         "
+PanDDA (panddas) installed into its own prefix with ccp4/9.0's ccp4-python.
+ccp4/9.0 is auto-loaded as a runtime dependency; the PYTHONPATH prepend below
+shadows the panddas copy bundled inside ccp4.
+Main entry point: pandda.analyse
+"
+
+prepend-path    PATH        $PREFIX/bin
+prepend-path    PYTHONPATH  $PREFIX/site-packages