feat: Track table generation with progress bars

[Matt711]

Closes #216

Adds progress bars via --progress flag. The basic design is
straightforward. I'm using a global thread-safe ProgressTracker
that holds a HashMap<Table, TableProgress>. Each TableProgress
contains counters for parts and buffers, plus an indicatif::ProgressBar
for visualizing the bars. The tracker is cloned to all tasks and they all
update the same ProgressTracker. Updates happen at two points:
after each buffer/chunk write and after each file completes. The MultiProgress
coordinator is stored to keep the display alive. The progress visualization
is disabled by default. tpchgen-rs is usually too fast to turn it on until
about sf10 from local macbook experiments. As @clflushopt said,

This seems like a good UX improvement, although being fast enough that a progress bar isn't necessary might be better 😅

So lets keep going 🚀

Screen.Recording.2026-01-13.at.10.43.42.AM.mov

[alamb]

Thank you @Matt711 -- this looks very cool. I am not sure I will have time to review it properly. The screen recording looks really nice

One question I had was what do you think about showing progress by default? It seems like this might interact weirdly with #225 from @kevinjqliu -- specifically logging)

[kevinjqliu]

Just merged #225 which defines the logging level based on the given param (default, --quiet, --verbose)

tpchgen-rs/tpchgen-cli/bin/main.rs

Lines 177 to 192 in b91de80

	// Configure logging
	if self.quiet {
	// Quiet mode: only show error-level logs
	env_logger::builder()
	.filter_level(LevelFilter::Error)
	.init();
	} else if self.verbose {
	env_logger::builder().filter_level(LevelFilter::Info).init();
	info!("Verbose output enabled (ignoring RUST_LOG environment variable)");
	} else {
	// Default: show warnings and errors, but respect RUST_LOG if set
	env_logger::builder()
	.filter_level(LevelFilter::Warn)
	.parse_default_env()
	.init();
	}

I think we should take this into consideration when rendering the process bar. i.e. turn progress bar off for --quiet

[alamb]

I think we should take this into consideration when rendering the process bar. i.e. turn progress bar off for --quiet

Yeah, I was thinking we could potentially go all the way and enable the progress bar by default -- though I am not sure how that would interact with the printing of the default logging

[alamb]

One thing some other users have asked for is to reuse some of the tpchgen-cli machinery -- I suspect they may have ideas about using the ProgressTracker, etc

What do you think about abstracting the interface out -- e.g. so this is like Box<dyn ProgressTracker> or something

A default implementaion using multi_progress seems good

[alamb]

why allow dead code?

[kevinjqliu]

Yeah, I was thinking we could potentially go all the way and enable the progress bar by default -- though I am not sure how that would interact with the printing of the default logging

I had some time to think about the different interactions when implementing --quiet. The conclusion I've came to is

• --quiet is explicit and absolute. The user specifies this to silence ALL output. We should not display a progress bar when --quiet is specified.
• --verbose is a request for more information. Internally, we raise the logging level to output more information. Progress bar should show for this case.
• For the "default" case, we can show the progress bar. I think its a nice improvement to the UI

LMK what you think!

[kevinjqliu]

i tested this out locally and found an edge case where skip generation (because the file already exists) has some weird interactions

➜  tpchgen-rs git:(fea/progress-tracker) ✗ cargo run --bin tpchgen-cli --release -- --tables region --format csv -P              
    Finished `release` profile [optimized] target(s) in 0.45s
     Running `target/release/tpchgen-cli --tables region --format csv -P`
region     [░░░░░░░░░░░░░░░░░░░░░░░░░░░░] Parts:   0/1    Buffers:     0   0%                                                                                                                               Info: ./region.csv already exists, skipping generation
region     [████████████████████████████] Parts:   1/1    Buffers:     0 100%                                                                                                                               

[alamb]

LMK what you think!

That sounds like a good idea to me