Move building of python arrow modules from pyarrow to apache-arrow. (#…

…30285)

A couple things here:
1. fix FTBFS state that apache-arrow was currently in.

  * need zstd-static because:

CMake Error at /usr/lib64/cmake/zstd/zstdTargets.cmake:107 (message):
         The imported target "zstd::libzstd_static" references the file
            "/usr/lib/libzstd.a" uses=cmake/configure
         but this file does not exist.  Possible reasons include:

2. cherry-pick a patch (from 2016) to rapidjson that allows a user of it
   (apache-arrow) to build with gcc-14.

3. Add environment variables to build so the python build modules
   have access.  I had hoped to localize these or use substitution
   in the environment block, but that is not supported.

   chainguard-dev/melange#1548
   chainguard-dev/melange#1402

4. drop unnecessary runtime dependency from libarrow to libarrow.

5. remove pyarrow.

6. bump the epoch on apache-arrow to 3 to be newer than pyarrow was.

7. provide multiple versions of the python modules, and make them
provide 'pyarrow' for backwards compat.

apache-arrow.yaml

@@ -1,11 +1,22 @@
 package:
   name: apache-arrow
   version: 17.0.0
-  epoch: 0
+  epoch: 2
   description: "multi-language toolbox for accelerated data interchange and in-memory processing"
   copyright:
     - license: Apache-2.0
 
+vars:
+  pypi-package: pyarrow
+
+data:
+  - name: py-versions
+    items:
+      3.10: '310'
+      3.11: '311'
+      3.12: '312'
+      3.13: '300'
+
 environment:
   contents:
     packages:
@@ -19,6 +30,7 @@ environment:
       - bzip2-dev
       - c-ares-dev
       - cmake
+      - cython
       - gflags-dev
       - glog-dev
       - grpc
@@ -33,6 +45,9 @@ environment:
       - protobuf
       - protobuf-dev
       - protoc
+      - py3-supported-build-base-dev
+      - py3-supported-cython
+      - py3-supported-numpy
       - rapidjson-dev
       - re2-dev
       - snappy-dev
@@ -44,6 +59,13 @@ environment:
       - xsimd-dev
       - zlib-dev
       - zstd-dev
+      - zstd-static
+  environment:
+    # These PYARROW_* and CMAKE_PREFIX_PATH are used by the python build.
+    PYARROW_PARALLEL: "4"
+    PYARROW_WITH_PARQUET: "1"
+    PYARROW_WITH_ACERO: "1"
+    CMAKE_PREFIX_PATH: "/home/build/melange-out/apache-arrow/usr"
 
 pipeline:
   - uses: git-checkout
@@ -103,6 +125,42 @@ pipeline:
   - uses: strip
 
 subpackages:
+  - range: py-versions
+    name: py${{range.key}}-${{vars.pypi-package}}
+    description: ${{vars.pypi-package}} installed for python${{range.key}}
+    dependencies:
+      provider-priority: ${{range.value}}
+      provides:
+        - py3-${{vars.pypi-package}}
+        - pyarrow
+      runtime:
+        - py${{range.key}}-numpy
+    pipeline:
+      - working-directory: /home/build/apache-arrow/python
+        pipeline:
+          - uses: py/pip-build-install
+            with:
+              python: python${{range.key}}
+      - uses: strip
+    test:
+      pipeline:
+        - uses: python/import
+          with:
+            python: python${{range.key}}
+            imports: |
+              import pyarrow
+              from pyarrow import acero
+              from pyarrow import parquet
+
+  - name: py3-supported-${{vars.pypi-package}}
+    description: meta package providing ${{vars.pypi-package}} for supported python versions.
+    dependencies:
+      runtime:
+        - py3.10-${{vars.pypi-package}}
+        - py3.11-${{vars.pypi-package}}
+        - py3.12-${{vars.pypi-package}}
+        - py3.13-${{vars.pypi-package}}
+
   - name: ${{package.name}}-dev
     description: "multi-language toolbox for accelerated data interchange and in-memory processing (development files)"
     pipeline:
@@ -117,9 +175,6 @@ subpackages:
       - runs: |
           mkdir -p ${{targets.subpkgdir}}/usr/lib
           cp -a ${{targets.destdir}}/usr/lib/libarrow.so.* ${{targets.subpkgdir}}/usr/lib
-    dependencies:
-      runtime:
-        - libarrow
 
   - name: "libarrow_acero"
     description: "multi-language toolbox for accelerated data interchange and in-memory processing (libarrow_acero library)"

rapidjson.yaml

@@ -1,7 +1,7 @@
 package:
   name: rapidjson
   version: 1.1.0
-  epoch: 2
+  epoch: 3
   description: Fast JSON parser/generator for C++ with both SAX/DOM style API
   copyright:
     - license: BSD-2-Clause
@@ -23,6 +23,10 @@ pipeline:
       expected-commit: f54b0e47a08782a6131cc3d60f94d038fa6e0a51
       repository: https://github.com/miloyip/rapidjson
       tag: v${{package.version}}
+      cherry-picks: |
+        # https://github.com/Tencent/rapidjson/issues/2277
+        master/862c39be371278a45a88d4d1d75164be57bb7e2d: fix build on gcc-14
+        master/3b2441b87f99ab65f37b141a7b548ebadb607b96: fix build on gcc-14
 
   - uses: patch
     with: