Some debugging

Change-Id: I47af06be9407c0b7eb39e0fb8cd37970e8eeeb67

Author: wesm

cpp/src/arrow/ipc/adapter.cc

@@ -165,6 +165,8 @@ class RecordBatchWriter {
 
       // The buffer might be null if we are handling zero row lengths.
       if (buffer) {
+        DCHECK(buffer->capacity() >= buffer->size());
+
         // We use capacity here, because size might not reflect the padding
         // requirements of buffers but capacity always should.
         size = buffer->capacity();

python/pyarrow/tests/test_ipc.py

@@ -60,9 +60,12 @@ def test_ipc_simple_roundtrip():
 
 def big_batch():
     df = pd.DataFrame(
-        np.random.randn(2**4, 2**23).T,
+        np.random.randn(2**4, 2**20).T,
         columns=[str(i) for i in range(2**4)]
     )
+
+    df = pd.concat([df] * 2 ** 3, ignore_index=True)
+
     return A.RecordBatch.from_pandas(df)
 
 
@@ -82,3 +85,6 @@ def read_file(source):
     reader = ipc.ArrowFileReader(source)
     return [reader.get_record_batch(i)
             for i in range(reader.num_record_batches)]
+
+batch = big_batch()
+mem = write_to_memory(batch)

python/src/pyarrow/common.h

@@ -120,6 +120,7 @@ class PYARROW_EXPORT NumPyBuffer : public arrow::Buffer {
 
     data_ = reinterpret_cast<const uint8_t*>(PyArray_DATA(arr_));
     size_ = PyArray_SIZE(arr_);
+    capacity_ = size_;
   }
 
   virtual ~NumPyBuffer() {