Compression CLI flag ineffective - overridden to ZSTD across multiple layers (SamToNTuple + RAMNTupleRecord)

[ZaGrayWolf]

Description

The samtoramntuple CLI accepts a compression algorithm argument (zstd, lz4, lzma), but the output size remains identical regardless of the selected option.

Reproduction

Commands:
./tools/samtoramntuple sample.sam zstd.ram zstd
./tools/samtoramntuple sample.sam lz4.ram lz4
./tools/samtoramntuple sample.sam lzma.ram lzma

Observed output sizes:
zstd.ram → 22 MB
lz4.ram → 22 MB
lzma.ram → 22 MB

Code Analysis

Compression parameters are initially passed correctly:

src/ramcore/SamToNTuple.cxx:
writeOptions.SetCompression(compression_algorithm);

However, compression is overridden in multiple places:

1. In SamToNTuple.cxx:
   writeOptions.SetCompression(ROOT::RCompressionSetting::EAlgorithm::kZSTD, 1);

2. In src/rntuple/RAMNTupleRecord.cxx:
   writeOptions.SetCompression(505); // ZSTD level 5

As a result, the CLI-selected compression algorithm is not respected, and all outputs are effectively compressed using ZSTD.

Expected Behavior

The compression algorithm provided via CLI should determine the compression used in the output file.

Actual Behavior

Compression is always ZSTD regardless of CLI input.

Environment

• macOS (Apple Silicon ARM64)
• Test dataset: ~85 MB SAM (~309k reads)

Suggested Fix

Remove or conditionally apply hardcoded compression overrides so that CLI arguments are respected consistently across layers.

[neo-0007]

From what i see : The samtoramntuple cli tool does not take the compression algorithm as an argument

ramtools/tools/samtoramntuple.cxx

Lines 9 to 14 in 5cec39c

	std::cout << "Usage: " << argv[0] << " <input.sam> [output]\n";
	std::cout << "Options:\n";
	std::cout << " -split Split by chromosome\n";
	std::cout << " -noindex Disable indexing\n";
	std::cout << " -illumina Use Illumina quality binning\n";
	std::cout << " -dropqual Drop quality scores\n";

, It takes in the input file name and output filename , along with some optional flags

And as you can see in the samtoramntuple function call :

ramtools/tools/samtoramntuple.cxx

Line 58 in 5cec39c

samtoramntuple(input, ramfile.c_str(), do_index, true, true, 505, quality_mode);

We directly pass in 505 ( ZSTD L5) for now , its independent of any command line argument. Maybe it would be a good addition to add a command line arg to set the compression algorithm.

But your finding that in some places the compression algorithm parameter is ignored and zstd is used directly is totally valid, here are the places i was able to track down :

This probably needs to be fixed to use the passed in parameter :

ramtools/src/ramcore/SamToNTuple.cxx

Line 230 in 5cec39c

writeOptions.SetCompression(ROOT::RCompressionSetting::EAlgorithm::kZSTD, 1);

These are the compression set for writing the refs and indexes not the actual data so maybe its alright here:

ramtools/src/rntuple/RAMNTupleRecord.cxx

Line 203 in 5cec39c

writeOptions.SetCompression(505);

ramtools/src/rntuple/RAMNTupleRecord.cxx

Line 260 in 5cec39c

writeOptions.SetCompression(505);

This code is probably not used by the samtoramntuple , since we do the conversion logic inside the above function , but according to me it still needs to be fixed

ramtools/src/rntuple/RAMNTupleRecord.cxx

Line 535 in 5cec39c

writeOptions.SetCompression(505); // ZSTD level 5

Hope this helps to work on your PR #50 , Thanks!