WIP: ARROW-1796: [Python/Parquet] RowGroup filtering on file level

cpp/src/arrow/python/inference.cc

@@ -317,7 +317,7 @@ class TypeInferrer {
 
     if (obj == Py_None || internal::PyFloat_IsNaN(obj)) {
       ++none_count_;
-    } else if (PyBool_Check(obj)) {
+    } else if (internal::PyBoolScalar_Check(obj)) {
       ++bool_count_;
       *keep_going = make_unions_;
     } else if (internal::PyFloatScalar_Check(obj)) {
@@ -583,5 +583,20 @@ Status InferArrowTypeAndSize(PyObject* obj, int64_t* size,
   return Status::OK();
 }
 
+ARROW_EXPORT
+bool IsPyBool(PyObject* obj) {
+  return internal::PyBoolScalar_Check(obj);
+}
+
+ARROW_EXPORT
+bool IsPyInt(PyObject* obj) {
+  return internal::PyIntScalar_Check(obj);
+}
+
+ARROW_EXPORT
+bool IsPyFloat(PyObject* obj) {
+  return internal::PyFloatScalar_Check(obj);
+}
+
 }  // namespace py
 }  // namespace arrow

cpp/src/arrow/python/inference.h

@@ -40,6 +40,15 @@ class Status;
 
 namespace py {
 
+ARROW_EXPORT
+bool IsPyBool(PyObject* obj);
+
+ARROW_EXPORT
+bool IsPyInt(PyObject* obj);
+
+ARROW_EXPORT
+bool IsPyFloat(PyObject* obj);
+
 // These three functions take a sequence input, not arbitrary iterables
 ARROW_EXPORT
 arrow::Status InferArrowType(PyObject* obj, std::shared_ptr<arrow::DataType>* out_type);

cpp/src/arrow/python/numpy-internal.h

@@ -165,6 +165,10 @@ inline bool PyIntScalar_Check(PyObject* obj) {
   return PyLong_Check(obj) || PyArray_IsScalar(obj, Integer);
 }
 
+inline bool PyBoolScalar_Check(PyObject* obj) {
+  return PyBool_Check(obj) || PyArray_IsScalar(obj, Bool);
+}
+
 }  // namespace internal
 
 }  // namespace py

python/pyarrow/filesystem.py

@@ -149,7 +149,8 @@ def _isfilestore(self):
 
     def read_parquet(self, path, columns=None, metadata=None, schema=None,
                      use_threads=True, nthreads=None,
-                     use_pandas_metadata=False):
+                     use_pandas_metadata=False, filters=None,
+                     exact_filter_evaluation=True):
         """
         Read Parquet data from path in file system. Can read from a single file
         or a directory of files
@@ -179,9 +180,10 @@ def read_parquet(self, path, columns=None, metadata=None, schema=None,
         from pyarrow.util import _deprecate_nthreads
         use_threads = _deprecate_nthreads(use_threads, nthreads)
         dataset = ParquetDataset(path, schema=schema, metadata=metadata,
-                                 filesystem=self)
+                                 filesystem=self, filters=filters)
         return dataset.read(columns=columns, use_threads=use_threads,
-                            use_pandas_metadata=use_pandas_metadata)
+                            use_pandas_metadata=use_pandas_metadata,
+                            exact_filter_evaluation=exact_filter_evaluation)
 
     def open(self, path, mode='rb'):
         """

python/pyarrow/includes/libarrow.pxd

@@ -1064,6 +1064,12 @@ cdef extern from 'arrow/python/config.h' namespace 'arrow::py':
     void set_numpy_nan(object o)
 
 
+cdef extern from 'arrow/python/inference.h' namespace 'arrow::py':
+    c_bool IsPyBool(object o)
+    c_bool IsPyInt(object o)
+    c_bool IsPyFloat(object o)
+
+
 cdef extern from 'arrow/python/benchmark.h' namespace 'arrow::py::benchmark':
     void Benchmark_PandasObjectIsNull(object lst) except *