Sinks Batching Delay is only applied for postgres sink

[0xgouda]

Describe the bug
Sinks max Batching delay for flushing the measurements is only applied to the postgres sink

To Reproduce
Steps to reproduce the behavior:

1. go run ./cmd/pgwatch --sources=postgres://postgres:postgres@localhost:5432/postgres --sink=rpc://localhost:1234 --batching-delay 5m --sink=jsonfile://file.json
2. use basic preset for the monitored database
3. the logs will show peridoical writes every 1 minute (metric interval for basic preset) without consideration for the provided batching delay

Screenshots

go run ./cmd/pgwatch --sources=postgres://postgres:postgres@localhost:5432/postgres --sink=rpc://localhost:1234 --batching-delay 5m --sink=jsonfile://file.json
2025-06-25 15:50:22.955 [INFO] [sources:1] sources refreshed
2025-06-25 15:50:22.957 [INFO] [metrics:74] [presets:16] metrics and presets refreshed
2025-06-25 15:50:22.968 [INFO] [source:books344] [recovery:false] Connect OK. Version: PostgreSQL 16.9 (Ubuntu 16.9-0ubuntu0.24.04.1) on x86_64-pc-linux-gnu, compiled by gcc (Ubuntu 13.3.0-6ubuntu2~24.04) 13.3.0, 64-bit
2025-06-25 15:50:22.968 [INFO] [source:books344] [metric:wal] [interval:60] starting gatherer
2025-06-25 15:50:22.968 [INFO] [source:books344] [metric:db_size] [interval:300] starting gatherer
2025-06-25 15:50:22.968 [INFO] [source:books344] [metric:db_stats] [interval:60] starting gatherer
2025-06-25 15:50:22.969 [INFO] [source:books344] [metric:wal] [rows:1] [cache:false] measurements fetched
2025-06-25 15:50:22.969 [INFO] [sink:jsonfile] [filename:file.json] [rows:1] [elapsed:111.749µs] measurements written
2025-06-25 15:50:22.979 [INFO] [source:books344] [metric:db_stats] [rows:1] [cache:false] measurements fetched
2025-06-25 15:50:22.979 [INFO] [sink:jsonfile] [filename:file.json] [rows:1] [elapsed:48.327µs] measurements written
2025-06-25 15:50:22.980 [INFO] [source:books344] [metric:db_size] [rows:1] [cache:false] measurements fetched
2025-06-25 15:50:22.980 [INFO] [sink:jsonfile] [filename:file.json] [rows:1] [elapsed:25.478µs] measurements written
2025-06-25 15:51:23.002 [INFO] [source:books344] [metric:wal] [rows:1] [cache:false] measurements fetched
2025-06-25 15:51:23.003 [INFO] [sink:jsonfile] [filename:file.json] [rows:1] [elapsed:127.333µs] measurements written
2025-06-25 15:51:23.005 [INFO] [source:books344] [metric:db_stats] [rows:1] [cache:false] measurements fetched
2025-06-25 15:51:23.005 [INFO] [sink:jsonfile] [filename:file.json] [rows:1] [elapsed:48.533µs] measurements written

Additional context
inpsecting the codebase you will notice that opts.BatchingDelay is only used in postgres sink

[pashagolub]

It's not a bug. It's by design. Not every sink created equal. We 100% don't want Prometheus to be involved in cache. It works differently. File sink is also responsible for own caching.

As an improvement maybe it's worth to remove/rename the command-line parameter and make it sink-specific

[0xgouda]

what about RPC sink should it have its own caching?

[pashagolub]

So the rule is every sink decides whether it wants caching or not. This is implementation details. For example, I believe, some sinks could benefit from caching.

But the problem with global caching is this way we disable realtime reactions by some implementations.

Do you think if we specify in the manual that --batching-delay is a sink-specific option, it will solve the confusion we have atm?