Merge pull request #1105 from cyfdecyf/arrow

arrow: various improvements and fixes.

Author: waruqi

packages/a/arrow/xmake.lua

@@ -1,4 +1,6 @@
 package("arrow")
+    -- For how to enable various features and build PyArrow python packages,
+    -- refer to this discussion https://github.com/xmake-io/xmake-repo/discussions/1106 
 
     set_homepage("https://arrow.apache.org/")
     set_description("Apache Arrow is a multi-language toolbox for accelerated data interchange and in-memory processing")
@@ -8,22 +10,26 @@ package("arrow")
              "https://github.com/apache/arrow.git")
     add_versions('7.0.0', '57e13c62f27b710e1de54fd30faed612aefa22aa41fa2c0c3bacd204dd18a8f3')
 
-    add_configs(csv,      {description = "CSV reader module", default = true, type = "boolean"})
-    add_configs(json,     {description = "JSON reader module", default = false, type = "boolean"})
-    add_configs(engine,   {description = "Build the Arrow Execution Engine", default = true, type = "boolean"})
-    add_configs(dataset,  {description = "Dataset API, implies the Filesystem API", default = true, type = "boolean"})
-    add_configs(orc,      {description = "Arrow integration with Apache ORC", default = false, type = "boolean"})
-    add_configs(parquet,  {description = "Apache Parquet libraries and Arrow integration", default = false, type = "boolean"})
-    add_configs(python,   {description = "Enable Python C++ integration library. Requires python and numpy (not managed by xrepo).", default = false, type = "boolean"})
+    add_configs("csv",      {description = "CSV reader module", default = true, type = "boolean"})
+    add_configs("json",     {description = "JSON reader module", default = false, type = "boolean"})
+    add_configs("engine",   {description = "Build the Arrow Execution Engine", default = true, type = "boolean"})
+    add_configs("dataset",  {description = "Dataset API, implies the Filesystem API", default = true, type = "boolean"})
+    add_configs("orc",      {description = "Arrow integration with Apache ORC", default = false, type = "boolean"})
+    add_configs("parquet",  {description = "Apache Parquet libraries and Arrow integration", default = false, type = "boolean"})
+    add_configs("plasma",   {description = "Plasma Shared Memory Object Store", default = false, type = "boolean"})
+    add_configs("python",   {description = "Enable Python C++ integration library. Requires python and numpy (not managed by xmake/xrepo).", default = false, type = "boolean"})
     -- Arrow uses vendored mimalloc and jemalloc. Do not add these two libraries to configdeps.
-    add_configs(mimalloc, {description = "Build the Arrow mimalloc-based allocator", default = true, type = "boolean"})
-    add_configs(jemalloc, {description = "Build the Arrow jemalloc-based allocator", default = false, type = "boolean"})
+    add_configs("mimalloc", {description = "Build the Arrow mimalloc-based allocator", default = true, type = "boolean"})
+    add_configs("jemalloc", {description = "Build the Arrow jemalloc-based allocator", default = false, type = "boolean"})
+    -- If true, arrow will look for shared libraries for third party dependency.
+    -- The pyarrow python package creates shared library that links in all necessary thirdparty static libraries.
+    add_configs("shared_dep", {description = "Use shared library for dependency", default = false, type = "boolean"})
 
     -- Some libraries are required for build with our default config settings.
     local configdeps = {
         re2 = "re2", utf8proc = "utf8proc",
         -- compression libraries
-        brotli = "brotli", bz2 = "bz2", snappy = "snappy", lz4 = "lz4", zlib = "zlib", zstd = "std",
+        brotli = "brotli", bz2 = "bzip2", snappy = "snappy", lz4 = "lz4", zlib = "zlib", zstd = "zstd",
     }
     for config, dep in pairs(configdeps) do
         add_configs(config, {description = "Enable " .. dep .. " support.", default = false, type = "boolean"})
@@ -38,6 +44,18 @@ package("arrow")
     end
 
     on_load(function (package)
+        if package:config("plasma") then
+            package:add("links", "plasma")
+            package:add("deps", "gflags")
+        end
+        if package:config("parquet") then
+            package:add("links", "parquet")
+        end
+        if package:config("dataset") then
+            package:add("links", "arrow_dataset")
+        end
+        package:add("links", "arrow", "arrow_bundled_dependencies")
+
         for name, dep in pairs(configdeps) do
             if package:config(name) then
                 package:add("deps", dep)
@@ -58,15 +76,6 @@ package("arrow")
             package:add("deps", "zstd")
         end
         if package:config("parquet") then
-            cprint([[
-${yellow}In case of boost dependency conflicts, please use following code (order is important):
-
-    local boost_config = {configs = {system = true}} -- change this according to what your need
-    add_requires("thrift")
-    add_requireconfs("thrift.boost", boost_config)
-    add_requires("arrow", {configs = {parquet = true}})
-    add_requireconfs("arrow.boost", boost_config)
-]])
             package:add("deps", "thrift", configs)
         end
     end)
@@ -82,7 +91,7 @@ ${yellow}In case of boost dependency conflicts, please use following code (order
         local shared = package:config("shared")
         table.insert(configs, "-DARROW_BUILD_STATIC=" .. (shared and "OFF" or "ON"))
         table.insert(configs, "-DARROW_BUILD_SHARED=" .. (shared and "ON" or "OFF"))
-        table.insert(configs, "-DARROW_DEPENDENCY_USE_SHARED=" .. (shared and "ON" or "OFF"))
+        table.insert(configs, "-DARROW_DEPENDENCY_USE_SHARED=" .. (package:config("shared_dep") and "ON" or "OFF"))
 
         for config, enabled in pairs(package:configs()) do
             if not package:extraconf("configs", config, "builtin") and configdeps[config] == nil then