holos-tda 0.9.0

Vietoris-Rips persistence and checked degree-Rips modules
Documentation 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155

use std::path::PathBuf;

use clap::{Parser, ValueEnum};

use crate::collapse::{
    CollapseObjective, CollapsePortfolioArtifact, CollapsePortfolioCandidate,
    CollapsePortfolioDecodeLimits, CollapsePortfolioLimits, CollapsePortfolioObjective,
    collapse_sparse_portfolio,
};
use crate::io;

use super::{version_string, write_via_temporary};

#[derive(Debug, Clone, Copy, PartialEq, Eq, ValueEnum)]
enum CandidateArg {
    Serial,
    Rounds,
    Adaptive,
}

#[derive(Debug, Clone, Copy, PartialEq, Eq, ValueEnum)]
enum ScoreArg {
    Edges,
    Columns,
}

#[derive(Parser)]
#[command(
    name = "holos collapse-portfolio",
    version = version_string(),
    about = "Select and certify the best declared collapse schedule"
)]
pub(super) struct CollapsePortfolioCli {
    /// Sparse `i j d` input.
    input: PathBuf,

    /// Output `HOLOSPOR` artifact.
    output: PathBuf,

    /// Schedules in tie-breaking order.
    #[arg(
        long = "candidate",
        value_enum,
        value_delimiter = ',',
        default_value = "serial,rounds,adaptive"
    )]
    candidates: Vec<CandidateArg>,

    /// Filtration threshold. The default includes every listed edge.
    #[arg(long, value_name = "T")]
    threshold: Option<f64>,

    /// Objective used to compare surviving graphs.
    #[arg(long, value_enum, default_value_t = ScoreArg::Columns)]
    score: ScoreArg,

    /// Highest homology dimension served by a column score.
    #[arg(long, value_name = "D", default_value_t = 1)]
    dim: usize,

    /// Worker count for the rounds schedule and input parser.
    #[arg(long, value_name = "N", default_value_t = 4)]
    threads: usize,

    /// Downstream objective for the adaptive schedule.
    #[arg(long, value_enum, default_value_t = super::ObjectiveArg::H1)]
    adaptive_objective: super::ObjectiveArg,

    /// Removability-test limit for the adaptive schedule.
    #[arg(long, value_name = "N")]
    adaptive_work_limit: Option<u64>,

    /// Largest nonvertex clique count visited for one schedule.
    #[arg(long, value_name = "N", default_value_t = 100_000_000)]
    max_cliques: u64,

    /// Largest accepted output artifact.
    #[arg(long, value_name = "BYTES", default_value_t = 1usize << 30)]
    max_artifact_bytes: usize,
}

pub(super) fn run_collapse_portfolio(cli: CollapsePortfolioCli) -> crate::Result<()> {
    let graph = io::read_sparse_matrix(&cli.input, cli.threads)?;
    let candidates = cli
        .candidates
        .iter()
        .map(|candidate| portfolio_candidate(*candidate, &cli))
        .collect::<Vec<_>>();
    let objective = match cli.score {
        ScoreArg::Edges => CollapsePortfolioObjective::Edges,
        ScoreArg::Columns => CollapsePortfolioObjective::ReductionColumns {
            max_homology_dimension: cli.dim,
        },
    };
    let limits = CollapsePortfolioLimits {
        max_candidates: candidates.len(),
        max_homology_dimension: cli.dim,
        max_cliques_per_candidate: cli.max_cliques,
    };
    let portfolio =
        collapse_sparse_portfolio(&graph, cli.threshold, &candidates, objective, limits)?;
    let artifact = CollapsePortfolioArtifact::from_portfolio(&portfolio, limits)?;
    let decode_limits = CollapsePortfolioDecodeLimits {
        max_bytes: cli.max_artifact_bytes,
        max_candidates: candidates.len(),
        ..CollapsePortfolioDecodeLimits::default()
    };
    let bytes = artifact.encode(limits, decode_limits)?;
    write_via_temporary(&cli.output, &bytes)?;
    print_portfolio(&portfolio, bytes.len());
    Ok(())
}

fn portfolio_candidate(
    candidate: CandidateArg,
    cli: &CollapsePortfolioCli,
) -> CollapsePortfolioCandidate {
    match candidate {
        CandidateArg::Serial => CollapsePortfolioCandidate::Serial,
        CandidateArg::Rounds => CollapsePortfolioCandidate::Rounds {
            threads: cli.threads,
        },
        CandidateArg::Adaptive => CollapsePortfolioCandidate::Adaptive {
            objective: match cli.adaptive_objective {
                super::ObjectiveArg::H1 => CollapseObjective::H1,
                super::ObjectiveArg::H2 => CollapseObjective::H2,
            },
            work_limit: cli.adaptive_work_limit,
        },
    }
}

fn print_portfolio(portfolio: &crate::collapse::CollapsePortfolio, bytes: usize) {
    for (index, entry) in portfolio.entries().iter().enumerate() {
        println!(
            "candidate {index}: {}, simplex counts {:?}, {} surviving edges",
            candidate_name(entry.candidate()),
            entry.score().simplex_counts(),
            entry.result().matrix.num_edges(),
        );
    }
    println!(
        "selected candidate {} ({}) and wrote {bytes} bytes",
        portfolio.selected_index(),
        candidate_name(portfolio.selected().candidate()),
    );
}

fn candidate_name(candidate: CollapsePortfolioCandidate) -> &'static str {
    match candidate {
        CollapsePortfolioCandidate::Serial => "serial",
        CollapsePortfolioCandidate::Rounds { .. } => "rounds",
        CollapsePortfolioCandidate::Adaptive { .. } => "adaptive",
    }
}