[C++][Parquet] Separate encoders and decoder

[pitrou]

trafficstars

Describe the enhancement requested

Currently, encoders and decoders are defined in a single file encoding.cc, which is quite large.

Given that their infrastructure is separate, it would probably make maintenance easier to split them into two C++ source files (for example encoder.cc and decoder.cc). We can add corresponding .h files, and also keep encoding.h for compatibility.

Component(s)

C++, Parquet

[pitrou]

Thoughts @felipecrv @mapleFU @wgtmac ?

[mapleFU]

I just check that there're few common constant used in encoding.cc, so maybe spliting them is not hard.

But I'm not sure about this. Current code is also ok to me

[felipecrv]

split them into two C++ source files

I'm in favor, but it will be 3 files (or .h/.cc pairs) because there has to be one for the shared funcitonality. One risk of the split is exposing the bits that are now fully private in encoding.cc.

[wgtmac]

Perhaps we can add a encoding_internal.h file to hold those private but common stuff?

[pitrou]

Perhaps we can add a encoding_internal.h file to hold those private but common stuff?

We can, though I doubt there's much shared functionality.

[changkhothuychung]

Hi, I am new to apache arrow and I would like to take this issue. Can this one be assigned to me?

[mapleFU]

You can type "take" to take the issue, and create a pull request named "GH-40154: [C++][Parquet] ..." when you finished @changkhothuychung

[kou]

Assigned!