Copy test data files to the binary directory

[duanmeng]

ParquetReaderTest depends on the test data files in directory examples.
It is nice to copy it into the binary directory during the CMake project loading.

[netlify]

✅ Deploy Preview for meta-velox canceled.

Name	Link
🔨 Latest commit	d9066db
🔍 Latest deploy log	https://app.netlify.com/sites/meta-velox/deploys/65c1e090742178000825f5cd

[kgpai]

I get what you are doing here, but I reckon this must be a common enough pattern that there must be a simpler , idiomatic way of doing this.

cc : @assignUser

[kgpai]

@duanmeng I reckon your original implementation was more idiomatic (or correct) -
file(COPY assets DESTINATION ${CMAKE_BINARY_DIR})

@assignUser let us know what you think.

[assignUser]

I don't have a lot of bw at the moment but reading up on this a bit I am pretty sure that the custom command just implements the same features that file(COPY) gives us as that will also only be executed if the files/dir don't exist or are older. And it is automatically using relative paths.

The only difference would probably be when installing targets (which we don't do yet) where the custom target would work but file(COPY) would not. But I am not entirely sure about that but can't look into it further at the moment. Though it is unlikely people will run tests outside of the build tree (I don't see us installing tests onto the system?)

So file(COPY) as the more concise version is probably the better choice for now/this case.

I do appreciate the contribution, great to see! Thanks @duanmeng

[duanmeng]

So file(COPY) as the more concise version is probably the better choice for now/this case.

@assignUser Thanks for your help. I've gone back to the file(COPY) way according to your and @kgpai advice :)

[facebook-github-bot]

@kgpai has imported this pull request. If you are a Meta employee, you can view this diff on Phabricator.

[facebook-github-bot]

@kgpai merged this pull request in db109f2.

[conbench-facebook]

Conbench analyzed the 1 benchmark run on commit db109f2f.

There were no benchmark performance regressions. 🎉

The full Conbench report has more details.