It's great to hear that you fixed the unit tests with 1.7.0.
Thank you for sharing. :)

Dongjoon.

On Wed, Nov 3, 2021 at 1:07 PM A L <anthonyn...@gmail.com> wrote:

> Hi Dongjoon,
>
> Thanks for the explanation.
> Yes, I agree that V1.2.3 is a very old version and it should have been
> upgraded long time ago.
> At the end I fix my unit test by providing only a temporary  File object
> without creating a real file in FileSystem.
> That's actually what my project's non-unit test code has already been
> doing.
>
> Anthony
>
>
>
>
>
> On Wed, Nov 3, 2021 at 12:30 PM Dongjoon Hyun <dongjoon.h...@gmail.com>
> wrote:
>
>> Hi, Anthony.
>>
>> Thank you for trying 1.7.0. It seems that your unit test reuses the test
>> file name.
>>
>> For breaking changes, I also raised similar breaking change issues at
>> 1.6.x and
>> fixed some in order to help the downstream migration.
>>
>>     TITLE: Apache ORC Versioning (Semantic Versioning)
>>     https://lists.apache.org/thread/nhw99jh1r0fc7r74cof0nhhdzvcqwvw5
>>
>> There is another recent discussion on the ORC releases too.
>>
>>     TITLE: [DISCUSS] Apache ORC Release Cadence
>>     https://lists.apache.org/thread/ql5o2ndon1b0818d4z5nb6001q09z5ck
>>
>> AFAIK, Apache ORC didn't follow Semantic Versioning officially until
>> 1.6.x.
>> We are still in the middle of transitioning toward `Semantic Versioning`
>> and enforcing it.
>>
>> BTW, you are talking about an ancient breaking change from 1.2.3
>> (2016-12-12) to 1.3.0 (2017-01-23).
>> They were archived a long time ago and recently Apache ORC 1.5 became EOL.
>> It could be an example, but it's beyond AS-IS scope of backward
>> compatibility at 1.6/1.7.
>>
>> Dongjoon.
>>
>>
>> On Tue, Nov 2, 2021 at 6:12 PM A L <anthonyn...@gmail.com> wrote:
>>
>>> Hi,
>>>
>>> My project used to use Apache orc-core 1.2.3, in my code I use
>>> createWriter(Path path, OrcFile.WriterOptions opts) to create a writer,
>>> path argument is a file exist in FileSystem. It worked well. After I
>>> upgraded orc-core to 1.7.0, I found that my unit test failed with the
>>> org.apache.hadoop.fs.FileAlreadyExistsException.
>>>
>>> I found that in ORC-119 there was a change that added a PhysicalFsWriter
>>> into WriterImpl constructor.
>>> In PhysicalFsWriter's constructor it checks if the file exist and if
>>> overwrite is true.
>>> So it's this change breaking the backward compatibility, has anyone had
>>> this issue when you upgrade the version?
>>> How did you fix it?
>>>
>>> Thanks,
>>>
>>> Anthony
>>>
>>

Reply via email to