[logachev]

We were using the same buffer size constant for Arrow and Json data retrieval. The problem was that while for JSON buffer represented rows, for Arrow it represented batches.

As a result, when using ReadAPI we allowed 20k batches which will be causing OOM or memory pressure issues when consumption is slower than fetching.

b/575989190

[gemini-code-assist]

Code Review

This pull request introduces a separate buffer size configuration for Arrow results in BigQueryStatement, replacing the use of getBufferSize() with a new getArrowBufferSize() method initialized to a default capacity of 32. A corresponding unit test was added to verify the bounded queue capacity. The feedback suggests changing the visibility of getArrowBufferSize() to private to improve encapsulation, as it is only used internally.

[gemini-code-assist]

The getArrowBufferSize() method is only used internally within the BigQueryStatement class. To adhere to encapsulation principles and minimize the class's API surface, its visibility should be restricted to private.

Suggested change

	int getArrowBufferSize() {
	return DEFAULT_ARROW_BUFFER_SIZE;
	}
	private int getArrowBufferSize() {
	return DEFAULT_ARROW_BUFFER_SIZE;
	}