This is an automated email from the ASF dual-hosted git repository.
wesm pushed a commit to branch master
in repository https://gitbox.apache.org/repos/asf/arrow.git
The following commit(s) were added to refs/heads/master by this push:
new 6fc9922 ARROW-2118: [C++] Fix misleading error when memory mapping a
zero-length file
6fc9922 is described below
commit 6fc9922a66017ab24f1a1676b83e4aaed0dd1930
Author: Wes McKinney <wes.mckin...@twosigma.com>
AuthorDate: Mon Mar 12 14:41:14 2018 -0400
ARROW-2118: [C++] Fix misleading error when memory mapping a zero-length
file
This was causing an unintuitive error to bubble up when reading a
malformed, zero-length Parquet file. To improve the actual Parquet error
message, we must complete PARQUET-1243
Author: Wes McKinney <wes.mckin...@twosigma.com>
Closes #1735 from wesm/ARROW-2118 and squashes the following commits:
dd5dcb06 <Wes McKinney> Use std::strerror to make errno more helpful
7717689d <Wes McKinney> Add Python unit test
10a368fc <Wes McKinney> Do not return error when memory-mapping length 0
file
---
cpp/src/arrow/io/file.cc | 20 +++++++++++++-------
cpp/src/arrow/io/io-file-test.cc | 10 ++++++++++
python/pyarrow/tests/test_io.py | 8 ++++++++
3 files changed, 31 insertions(+), 7 deletions(-)
diff --git a/cpp/src/arrow/io/file.cc b/cpp/src/arrow/io/file.cc
index d44d90c..02cc4db 100644
--- a/cpp/src/arrow/io/file.cc
+++ b/cpp/src/arrow/io/file.cc
@@ -624,16 +624,22 @@ class MemoryMappedFile::MemoryMap : public MutableBuffer {
is_mutable_ = false;
}
- void* result = mmap(nullptr, static_cast<size_t>(file_->size()),
prot_flags, map_mode,
- file_->fd(), 0);
- if (result == MAP_FAILED) {
- std::stringstream ss;
- ss << "Memory mapping file failed, errno: " << errno;
- return Status::IOError(ss.str());
+ size_ = file_->size();
+
+ void* result = nullptr;
+
+ // Memory mapping fails when file size is 0
+ if (size_ > 0) {
+ result =
+ mmap(nullptr, static_cast<size_t>(size_), prot_flags, map_mode,
file_->fd(), 0);
+ if (result == MAP_FAILED) {
+ std::stringstream ss;
+ ss << "Memory mapping file failed: " << std::strerror(errno);
+ return Status::IOError(ss.str());
+ }
}
data_ = mutable_data_ = reinterpret_cast<uint8_t*>(result);
- size_ = file_->size();
position_ = 0;
diff --git a/cpp/src/arrow/io/io-file-test.cc b/cpp/src/arrow/io/io-file-test.cc
index 2a4acab..53218ca 100644
--- a/cpp/src/arrow/io/io-file-test.cc
+++ b/cpp/src/arrow/io/io-file-test.cc
@@ -467,6 +467,16 @@ class TestMemoryMappedFile : public ::testing::Test,
public MemoryMapFixture {
TEST_F(TestMemoryMappedFile, InvalidUsages) {}
+TEST_F(TestMemoryMappedFile, ZeroSizeFlie) {
+ std::string path = "io-memory-map-zero-size";
+ std::shared_ptr<MemoryMappedFile> result;
+ ASSERT_OK(InitMemoryMap(0, path, &result));
+
+ int64_t size = 0;
+ ASSERT_OK(result->Tell(&size));
+ ASSERT_EQ(0, size);
+}
+
TEST_F(TestMemoryMappedFile, WriteRead) {
const int64_t buffer_size = 1024;
std::vector<uint8_t> buffer(buffer_size);
diff --git a/python/pyarrow/tests/test_io.py b/python/pyarrow/tests/test_io.py
index 5913810..b29b9f1 100644
--- a/python/pyarrow/tests/test_io.py
+++ b/python/pyarrow/tests/test_io.py
@@ -592,6 +592,14 @@ def test_memory_map_writer(tmpdir):
assert f.read(3) == b'foo'
+def test_memory_zero_length(tmpdir):
+ path = os.path.join(str(tmpdir), guid())
+ f = open(path, 'wb')
+ f.close()
+ with pa.memory_map(path, mode='r+b') as memory_map:
+ assert memory_map.size() == 0
+
+
def test_os_file_writer(tmpdir):
SIZE = 4096
arr = np.random.randint(0, 256, size=SIZE).astype('u1')
--
To stop receiving notification emails like this one, please contact
w...@apache.org.
