There is no failed activity log when previewing one invalid avro file

[v-xianya]

Storage Explorer Version: 1.33.0-dev (96)
Build Number: 20231103.1
Branch: main
Platform/OS: Windows 10/Linux Ubuntu 22.04/MacOS Sonoma 14.1 (Apple M1 Pro)
Architecture: x64/x64/arm64
How Found: From running test cases
Regression From: Previous release (1.32.0)

Steps to Reproduce

1. Expand one storage account -> Blob Containers.
2. Create a blob container -> Upload an invalid avro file.
3. Select the file -> Click 'Preview'.
4. Check whether there is a failed activity log.

Expected Experience

There is a failed activity log.

Actual Experience

There is no activity log.

[craxal]

@v-xianya I changed the library we use to parse Avro data in order to get better support for date formatting. There are a couple of things I've discovered:

• The library throws an error if there's a problem with the header data (missing magic bits, bad schema, etc.)
• The library does not throw an error if the invalid data is at the end of the file.

Can you:

1. Share the file that should be resulting in a failed activity.
2. If there is no error, what does the preview tab actually show?
3. Describe the contents of the file and how the data was generated.
   • What parts of the data are valid? What parts are invalid?
   • Is the schema at the head of the file able to be parsed? Is the invalid part of the data at the end of the file?

It's possible the newer library is more robust than the previous one, which would mean fewer errors.

[v-xianya]

@craxal I shared the invalid avro file on teams.
The preview tab screenshot as below:

[craxal]

@v-xianya Some changes were made recently. Does this still repro?

[v-xianya]

Hi @craxal Verified this issue on the main build 20240130.14. It reproduces.
.

[craxal]

After debugging into the Avro parsing library, it would seem that the invalid Avro file has a bad header. I found that the library mysteriously does not throw errors in this case. As a result, we simply don't get a schema object, and an empty set of columns and rows is returned.

To work around this issue, I think it's best to throw our own error. If the schema remains undefined when the library returns, it's likely the header was invalid, and so an error would be appropriate.

I've opened an issue for the library: mtth/avsc#448.

[v-xianya]

Verified this issue on the main build 20240201.4. Fixed.