Graphs

[fsktom]

I've been procrastinating on this for so long. It's possibly the most important aspect of this program. It's the reason I created this - Python was so slow for creating relative (% of plays at each time) graphs...

It's been 11 months since I started this Rust project, time to include graphs :)

[fsktom]

rusty-endsong-parser/stuff/ideas.md

Lines 41 to 44 in 57be173

	- plotting
	- either "plotpy - crates.io: Rust Package Registry" <https://crates.io/crates/plotpy>
	- or automatically call a Python script on your own <https://www.reddit.com/r/rust/comments/8h22h3/graphing_in_rust/dygs2xt?utm_medium=android_app&utm_source=share&context=3>
	- see [this repo](https://github.com/bheisler/cargo-criterion) for `gnuplot` and `plotters` usage

rusty-endsong-parser/stuff/ideas.md

Line 58 in 57be173

- [Plotters](https://old.reddit.com/r/rust/comments/ude3lz/plotters_is_back/) for graphs?

[fsktom]

the Python version uses matplotlib https://github.com/fsktom/endsong-parser-python

[fsktom]

probably plotly.rs
https://lib.rs/crates/plotly

I like that plot.show(); would open the default web browser (open HTML file). I could create HTML files for every graph created.
e.g.
folder plots/
with files <info_about_plot> - <time>.html hmm

[fsktom]

even if plotters is used by FAAAR more crates and seems more flexible

but probably more difficult to get in to

[fsktom]

Use multiple plotly traces for a compare command? (e.g. compare two artists)

[fsktom]

I love the sensible defaults in plotly. I don't have to manually scald the graph or set the dimensions <3

[fsktom]

https://github.com/flamegraph-rs/flamegraph#systems-performance-work-guided-by-flamegraphs for performance profiling

[fsktom]

W T F
ChatGPT reduced my code that runs (in release mode) for about 30s to about 5s wtffff
this function is called very often

fn gather_artist_date(
    entries: &Vec<SongEntry>,
    art: &Artist,
    start: &DateTime<Tz>,
    end: &DateTime<Tz>,
) -> u32 {
    let mut plays = 0;

    for entry in entries {
        let artist = Artist::from_str(&entry.artist);

        if artist == *art && entry.timestamp.ge(start) && entry.timestamp.le(end) {
            plays += 1;
        }
    }

    plays
}

ChatGPT recommendation with some changes of mine ->

fn gather_artist_date(
    entries: &[SongEntry],
    art: &Artist,
    start: &DateTime<Tz>,
    end: &DateTime<Tz>,
) -> usize {
    entries
        .iter()
        .filter(|entry| {
            entry.artist.eq(&art.name) && entry.timestamp.ge(start) && entry.timestamp.le(end)
        })
        .count()
}

[fsktom]

Ok, it seems like constantly creating a new Artist was the reason for the low performance and the functional-style iterator thingy did not improve performance (very likely due to compiler optimizations with --release), but comparing Strings directly without reallocating them.

[fsktom]

The problem will be then more clutter with comparing albums and esp. songs

[fsktom]

I'm so happy that I got it working 😄

[fsktom]

plot.show() doesn't work in Windows (also WSL) :))))
plotly/plotly.rs#132

[fsktom]

The problem will be then more clutter with comparing albums and esp. songs

such an easy fix....
I already had the Music trait.... just add a function to compare to a SongEntry :)
done in 5cb091d

[fsktom]

ehm... on my MacBook... it's basically instant - less than a second I think
doing something like (in release mode)

>>> gartr
Artist name?
  >> Freedom cALL

w t f

[fsktom]

Why is WSL performance sooo bad. And the input delay while typing...

[fsktom]

See #23 for plot comparison