Add python support for explain, which returns LLVM IR for the query

Author: alexbaden

python/pyhdk/_sql.pxd

@@ -73,6 +73,7 @@ cdef extern from "omniscidb/QueryEngine/Descriptors/RelAlgExecutionDescriptor.h"
 
     const CResultSetPtr& getRows()
     const vector[CTargetMetaInfo]& getTargetsMeta()
+    string getExplanation()
 
 cdef extern from "omniscidb/QueryEngine/RelAlgExecutor.h":
   cdef cppclass CRelAlgExecutor "RelAlgExecutor":

python/pyhdk/_sql.pyx

@@ -74,6 +74,9 @@ cdef class ExecutionResult:
     cdef shared_ptr[CArrowTable] at = converter.get().convertToArrowTable()
     return pyarrow_wrap_table(at)
 
+  def to_explain_str(self):
+    return self.c_result.getExplanation()
+
 cdef class RelAlgExecutor:
   cdef shared_ptr[CRelAlgExecutor] c_rel_alg_executor
   # DataMgr is used only to pass it to each produced ExecutionResult
@@ -108,6 +111,7 @@ cdef class RelAlgExecutor:
     c_eo.output_columnar_hint = kwargs.get("enable_columnar_output", g_enable_columnar_output)
     c_eo.with_watchdog = kwargs.get("enable_watchdog", g_enable_watchdog)
     c_eo.with_dynamic_watchdog = kwargs.get("enable_dynamic_watchdog", g_enable_dynamic_watchdog)
+    c_eo.just_explain = kwargs.get("just_explain", False)
     cdef CExecutionResult c_res = self.c_rel_alg_executor.get().executeRelAlgQuery(c_co, c_eo, False)
     cdef ExecutionResult res = ExecutionResult()
     res.c_result = move(c_res)

python/tests/test_pyhdk_sql.py

@@ -29,16 +29,21 @@ def teardown_class(cls):
         del cls.storage
 
     @classmethod
-    def execute_sql(cls, sql):
+    def execute_sql(cls, sql, **kwargs):
         ra = cls.calcite.process(sql)
         rel_alg_executor = pyhdk.sql.RelAlgExecutor(
             cls.executor, cls.storage, cls.data_mgr, ra
         )
-        return rel_alg_executor.execute()
+        return rel_alg_executor.execute(**kwargs)
 
     def test_simple_projection(self):
         res = self.execute_sql("SELECT * FROM test;")
         df = res.to_arrow().to_pandas()
         assert df.shape == (3, 2)
         assert df["a"].tolist() == [1, 2, 3]
         assert df["b"].tolist() == [10, 20, 30]
+
+    def test_explain(self):
+        res = self.execute_sql("SELECT * FROM test;", just_explain = True)
+        explain_str = res.to_explain_str()
+        assert (explain_str[:15] == "IR for the CPU:")