fallow_engine/

baseline_growth.rs

//! The shrink-only rule for a committed baseline.
//!
//! A baseline records debt that a repository still must pay. A change that adds
//! a finding can also re-save the baseline, and then the new finding is
//! suppressed and no stale entry exists. This module compares the baseline file
//! with the same file at a base ref and names each key that the base file does
//! not have. The comparison reads two file versions only. It needs no analysis
//! run, so it gives the same answer on a whole-project run and on a narrowed run.
//!
//! A key is compared as the file writes it. A renamed file gives a new key, so
//! the rule counts it as growth. The rule is strict on purpose: a reviewer
//! approves each new key, or the change removes it. A dead-code baseline with
//! an `identity` stores line-free keys once for each occurrence, so a moved
//! line is not growth and one more occurrence of a key is.

use std::collections::{BTreeMap, BTreeSet};
use std::path::Path;

use serde_json::{Map, Value};

use fallow_types::identity::{IdentityPaths, dead_code_canonical_key};

use crate::baseline::BaselineKind;

/// The keys that a baseline has and its version at the base ref does not have.
#[derive(Debug, Clone, Default, PartialEq, Eq)]
pub struct BaselineGrowth {
    /// One item per category that grew, sorted by category name.
    pub categories: Vec<GrownCategory>,
}

/// The new keys of one baseline category.
#[derive(Debug, Clone, PartialEq, Eq)]
pub struct GrownCategory {
    /// The category as the baseline file names it, for example `unused_files`.
    pub category: String,
    /// The new keys, sorted.
    pub keys: Vec<String>,
}

impl BaselineGrowth {
    /// The number of new keys in all categories.
    #[must_use]
    pub fn added_entries(&self) -> usize {
        self.categories.iter().map(|grown| grown.keys.len()).sum()
    }

    /// True when the baseline has no key that the base does not have.
    #[must_use]
    pub fn is_empty(&self) -> bool {
        self.categories.is_empty()
    }
}

/// Compare a baseline (`head`) with its version at the base ref (`base`).
///
/// Both values are the parsed JSON files. `kind` selects the categories that
/// the format of the command writes:
///
/// - `dead-code`: every top-level array of keys. When both files carry the
///   same `identity`, each extra occurrence of a key is growth. When the change
///   rewrote a legacy baseline, each legacy key is translated to its canonical
///   key and compared the same way. A category that holds a legacy key with no
///   unambiguous translation (a key with a line, a bare package name, a key
///   that lacks a part of the canonical key) is compared by its entry count.
///   Two legacy files compare as key sets.
/// - `dupes`: the content fingerprints of the clone groups. A base file saved
///   before fingerprints existed is compared by its `clone_groups` keys.
/// - `health`: the legacy `findings` keys, the finding counts per file and
///   category (a higher count is growth), the runtime-coverage finding IDs and
///   the refactoring target keys. When both files carry per-function counts,
///   those counts replace the per-file counts.
#[must_use]
pub fn baseline_growth(kind: BaselineKind, base: &Value, head: &Value) -> BaselineGrowth {
    let empty = Map::new();
    let base = base.as_object().unwrap_or(&empty);
    let head = head.as_object().unwrap_or(&empty);
    let mut grown: BTreeMap<String, Vec<String>> = BTreeMap::new();
    match kind {
        BaselineKind::DeadCode => {
            let base_scheme = base.get(DEAD_CODE_IDENTITY);
            let head_scheme = head.get(DEAD_CODE_IDENTITY);
            for (category, value) in head {
                if !value.is_array() {
                    continue;
                }
                if base_scheme.is_none() && head_scheme.is_some() {
                    add_grown_after_upgrade(&mut grown, category, base, head);
                } else if base_scheme != head_scheme {
                    add_grown_entry_count(&mut grown, category, base, head);
                } else if head_scheme.is_some() {
                    add_new_occurrences(&mut grown, category, base, head);
                } else {
                    add_new_keys(&mut grown, category, base, head);
                }
            }
        }
        BaselineKind::Dupes => {
            let category = if base.contains_key(DUPES_FINGERPRINTS) {
                DUPES_FINGERPRINTS
            } else {
                DUPES_LEGACY_GROUPS
            };
            add_new_keys(&mut grown, category, base, head);
        }
        BaselineKind::Health => {
            for category in HEALTH_KEY_CATEGORIES {
                add_new_keys(&mut grown, category, base, head);
            }
            let counts = if has_entries(base, HEALTH_IDENTITY_COUNTS)
                && has_entries(head, HEALTH_IDENTITY_COUNTS)
            {
                HEALTH_IDENTITY_COUNTS
            } else {
                HEALTH_FILE_COUNTS
            };
            add_grown_counts(&mut grown, counts, base, head);
        }
    }
    BaselineGrowth {
        categories: grown
            .into_iter()
            .filter(|(_, keys)| !keys.is_empty())
            .map(|(category, keys)| GrownCategory { category, keys })
            .collect(),
    }
}

const DEAD_CODE_IDENTITY: &str = "identity";
const DUPES_FINGERPRINTS: &str = "normalized_clone_fingerprints";
const DUPES_LEGACY_GROUPS: &str = "clone_groups";
const HEALTH_KEY_CATEGORIES: [&str; 3] = ["findings", "runtime_coverage_findings", "target_keys"];
const HEALTH_FILE_COUNTS: &str = "finding_counts";
const HEALTH_IDENTITY_COUNTS: &str = "identity_finding_counts";

/// The string members of the array `category`, as a set.
fn key_set<'a>(object: &'a Map<String, Value>, category: &str) -> BTreeSet<&'a str> {
    object
        .get(category)
        .and_then(Value::as_array)
        .map(|keys| keys.iter().filter_map(Value::as_str).collect())
        .unwrap_or_default()
}

fn has_entries(object: &Map<String, Value>, category: &str) -> bool {
    object
        .get(category)
        .and_then(Value::as_object)
        .is_some_and(|buckets| !buckets.is_empty())
}

fn add_new_keys(
    grown: &mut BTreeMap<String, Vec<String>>,
    category: &str,
    base: &Map<String, Value>,
    head: &Map<String, Value>,
) {
    let known = key_set(base, category);
    let added: Vec<String> = key_set(head, category)
        .into_iter()
        .filter(|key| !known.contains(key))
        .map(display_key)
        .collect();
    if !added.is_empty() {
        grown.entry(category.to_owned()).or_default().extend(added);
    }
}

/// The string members of the array `category`, with the number of times each
/// occurs.
fn key_counts<'a>(object: &'a Map<String, Value>, category: &str) -> BTreeMap<&'a str, usize> {
    let mut counts = BTreeMap::new();
    for key in object
        .get(category)
        .and_then(Value::as_array)
        .into_iter()
        .flatten()
        .filter_map(Value::as_str)
    {
        *counts.entry(key).or_insert(0) += 1;
    }
    counts
}

/// One entry for each occurrence of a key beyond its count at the base.
fn add_new_occurrences(
    grown: &mut BTreeMap<String, Vec<String>>,
    category: &str,
    base: &Map<String, Value>,
    head: &Map<String, Value>,
) {
    let known: BTreeMap<String, usize> = key_counts(base, category)
        .into_iter()
        .map(|(key, count)| (key.to_owned(), count))
        .collect();
    push_new_occurrences(grown, category, &known, head);
}

/// Push one entry for each occurrence of a head key beyond its `known` count.
fn push_new_occurrences(
    grown: &mut BTreeMap<String, Vec<String>>,
    category: &str,
    known: &BTreeMap<String, usize>,
    head: &Map<String, Value>,
) {
    let mut added = Vec::new();
    for (key, now) in key_counts(head, category) {
        let before = known.get(key).copied().unwrap_or(0);
        added.extend(std::iter::repeat_n(
            display_key(key),
            now.saturating_sub(before),
        ));
    }
    if !added.is_empty() {
        grown.entry(category.to_owned()).or_default().extend(added);
    }
}

/// Compare a legacy base category with a canonical head category. Every legacy
/// key must translate, or the category falls back to its entry count.
fn add_grown_after_upgrade(
    grown: &mut BTreeMap<String, Vec<String>>,
    category: &str,
    base: &Map<String, Value>,
    head: &Map<String, Value>,
) {
    let mut known: BTreeMap<String, usize> = BTreeMap::new();
    for (key, count) in key_counts(base, category) {
        let Some(canonical) = canonical_from_legacy(category, key) else {
            add_grown_entry_count(grown, category, base, head);
            return;
        };
        *known.entry(canonical).or_insert(0) += count;
    }
    push_new_occurrences(grown, category, &known, head);
}

/// The canonical key of a legacy dead-code baseline key, when the legacy form
/// holds every part of the canonical key and no line.
fn canonical_from_legacy(category: &str, key: &str) -> Option<String> {
    let paths = IdentityPaths::new(Path::new(""));
    let path = |value: &str| paths.key(Path::new(value));
    let set = |values: &[&str]| paths.set(values.iter().map(Path::new));
    let canonical = |rule: &str, parts: &[String]| {
        let parts: Vec<&str> = parts.iter().map(String::as_str).collect();
        Some(dead_code_canonical_key(rule, &parts))
    };
    let path_name = |rule: &str| {
        let (file, name) = key.split_once(':')?;
        canonical(rule, &[path(file), name.to_owned()])
    };
    let manifest_package = |rule: &str| {
        let (manifest, package) = key.rsplit_once(':')?;
        if !manifest.ends_with("package.json") {
            return None;
        }
        canonical(rule, &[path(manifest), package.to_owned()])
    };
    let member = |rule: &str| {
        let (file, member) = key.split_once(':')?;
        let (parent, name) = member.split_once('.')?;
        canonical(rule, &[path(file), parent.to_owned(), name.to_owned()])
    };
    match category {
        "unused_files" => canonical("unused-file", &[path(key)]),
        "unused_exports" => path_name("unused-export"),
        "unused_types" => path_name("unused-type"),
        "deprecated_exports_in_use" => path_name("deprecated-export-in-use"),
        "invalid_client_exports" => path_name("invalid-client-export"),
        "unresolved_imports" => path_name("unresolved-import"),
        "unprovided_injects" => path_name("unprovided-inject"),
        "unrendered_components" => path_name("unrendered-component"),
        "unused_server_actions" => path_name("unused-server-action"),
        "unused_load_data_keys" => path_name("unused-load-data-key"),
        "route_collisions" => path_name("route-collision"),
        "dynamic_segment_name_conflicts" => path_name("dynamic-segment-name-conflict"),
        "boundary_call_violations" => path_name("boundary-call-violation"),
        "boundary_coverage_violations" => canonical("boundary-coverage", &[path(key)]),
        "unused_dependencies" => manifest_package("unused-dependency"),
        "unused_dev_dependencies" => manifest_package("unused-dev-dependency"),
        "unused_optional_dependencies" => manifest_package("unused-optional-dependency"),
        "type_only_dependencies" => manifest_package("type-only-dependency"),
        "test_only_dependencies" => manifest_package("test-only-dependency"),
        "dev_dependencies_in_production" => manifest_package("dev-dependency-in-production"),
        "unused_enum_members" => member("unused-enum-member"),
        "unused_class_members" => member("unused-class-member"),
        "unused_store_members" => member("unused-store-member"),
        "unlisted_dependencies" => canonical("unlisted-dependency", &[key.to_owned()]),
        "private_type_leaks" => {
            let (file, rest) = key.split_once(':')?;
            let (export, leaked) = rest.split_once("->")?;
            canonical(
                "private-type-leak",
                &[path(file), export.to_owned(), leaked.to_owned()],
            )
        }
        "duplicate_exports" => {
            let mut parts = key.split('|');
            let name = parts.next()?.to_owned();
            let files: Vec<&str> = parts.collect();
            if files.is_empty() {
                return None;
            }
            canonical("duplicate-export", &[name, set(&files)])
        }
        "circular_dependencies" => {
            let files: Vec<&str> = key.split("->").collect();
            canonical("circular-dependency", &[set(&files)])
        }
        "re_export_cycles" => {
            let (kind, rest) = key.split_once(':')?;
            let files: Vec<&str> = rest.split("<->").collect();
            canonical("re-export-cycle", &[kind.to_owned(), set(&files)])
        }
        "boundary_violations" => {
            let (from, to) = key.split_once("->")?;
            canonical("boundary-violation", &[path(from), path(to)])
        }
        "unused_dependency_overrides" => {
            let (source, raw_key) = key.split_once(':')?;
            canonical(
                "unused-dependency-override",
                &[source.to_owned(), raw_key.to_owned()],
            )
        }
        "misconfigured_dependency_overrides" => {
            let (source, raw_key) = key.split_once(':')?;
            canonical(
                "misconfigured-dependency-override",
                &[source.to_owned(), raw_key.to_owned()],
            )
        }
        // Keys with a line (stale suppressions, misplaced directives, catalog
        // references), keys without the path or component name of the
        // canonical key (catalog entries and groups, component contracts),
        // keys whose parts cannot be split safely (policy violations) and
        // keys whose canonical key has fewer parts (client and server barrels)
        // have no translation.
        _ => None,
    }
}

/// One entry for each entry of `category` beyond its count at the base. Used
/// when the two files write keys in different forms.
fn add_grown_entry_count(
    grown: &mut BTreeMap<String, Vec<String>>,
    category: &str,
    base: &Map<String, Value>,
    head: &Map<String, Value>,
) {
    let entries = |object: &Map<String, Value>| {
        object
            .get(category)
            .and_then(Value::as_array)
            .map_or(0, Vec::len)
    };
    let (before, now) = (entries(base), entries(head));
    let extra = now.saturating_sub(before);
    let added = (1..=extra).map(|index| {
        format!(
            "new entry {index} of {extra}: the key format changed, so the gate compares entry counts ({before} -> {now})"
        )
    });
    if extra > 0 {
        grown.entry(category.to_owned()).or_default().extend(added);
    }
}

/// Health count buckets: `path -> finding category -> { count }`. A bucket
/// whose count is higher than at the base is growth, also when the base has no
/// such bucket.
fn add_grown_counts(
    grown: &mut BTreeMap<String, Vec<String>>,
    category: &str,
    base: &Map<String, Value>,
    head: &Map<String, Value>,
) {
    let Some(head_buckets) = head.get(category).and_then(Value::as_object) else {
        return;
    };
    let base_buckets = base.get(category).and_then(Value::as_object);
    let mut added = Vec::new();
    for (path, finding_counts) in head_buckets {
        let Some(finding_counts) = finding_counts.as_object() else {
            continue;
        };
        for (finding, count) in finding_counts {
            let now = bucket_count(Some(count));
            let before = bucket_count(
                base_buckets
                    .and_then(|buckets| buckets.get(path))
                    .and_then(|counts| counts.get(finding)),
            );
            if now > before {
                let key = display_key(path);
                added.push(if before == 0 {
                    format!("{key} {finding} (count {now})")
                } else {
                    format!("{key} {finding} (count {before} -> {now})")
                });
            }
        }
    }
    if !added.is_empty() {
        grown.entry(category.to_owned()).or_default().extend(added);
    }
}

fn bucket_count(bucket: Option<&Value>) -> u64 {
    bucket
        .and_then(|bucket| bucket.get("count"))
        .and_then(Value::as_u64)
        .unwrap_or(0)
}

/// Some formats join key parts with a NUL byte. Show it as `:`, so the
/// message stays on one line and prints in every terminal.
fn display_key(key: &str) -> String {
    key.replace('\0', ":")
}

/// Why the baseline at the base ref could not be read.
#[derive(Debug, Clone, PartialEq, Eq)]
pub enum BaseBaselineError {
    /// `git` could not be started.
    GitMissing(String),
    /// The baseline is not in a git work tree.
    NotARepository,
    /// Git cannot resolve the ref to a commit, typically because a shallow
    /// clone did not fetch it.
    RefUnavailable,
    /// Git failed for another reason.
    GitFailed(String),
}

/// Read the version of the file at `path` that the commit `git_ref` has.
///
/// `Ok(None)` when the commit has no file at that path, which is a baseline
/// that the change adds. The path is resolved relative to the directory of the
/// file, so a baseline in a nested directory or under `--root` reads correctly.
pub fn read_baseline_at_ref(
    path: &Path,
    git_ref: &str,
) -> Result<Option<String>, BaseBaselineError> {
    let dir = match path.parent() {
        Some(parent) if !parent.as_os_str().is_empty() => parent,
        _ => Path::new("."),
    };
    let Some(name) = path.file_name().and_then(|name| name.to_str()) else {
        return Err(BaseBaselineError::GitFailed(format!(
            "the baseline path {} has no file name",
            path.display()
        )));
    };
    let inside = run_git(dir, &["rev-parse", "--is-inside-work-tree"])?;
    if !inside.status.success() {
        return Err(BaseBaselineError::NotARepository);
    }
    let resolved = run_git(
        dir,
        &["rev-parse", "--verify", "--quiet", &peel_to_commit(git_ref)],
    )?;
    if !resolved.status.success() {
        return Err(BaseBaselineError::RefUnavailable);
    }
    let object = format!("{git_ref}:./{name}");
    let exists = run_git(dir, &["cat-file", "-e", &object])?;
    if !exists.status.success() {
        return Ok(None);
    }
    let shown = run_git(dir, &["show", &object])?;
    if !shown.status.success() {
        return Err(BaseBaselineError::GitFailed(
            String::from_utf8_lossy(&shown.stderr).trim().to_owned(),
        ));
    }
    Ok(Some(String::from_utf8_lossy(&shown.stdout).into_owned()))
}

/// True when `git_ref` names the commit that `HEAD` names in the repository
/// at `dir`. A base that resolves to `HEAD` compares a committed baseline with
/// itself, so only an uncommitted change can grow it.
#[must_use]
pub fn ref_is_head(dir: &Path, git_ref: &str) -> bool {
    let resolve = |rev: &str| {
        run_git(
            dir,
            &["rev-parse", "--verify", "--quiet", &peel_to_commit(rev)],
        )
        .ok()
        .filter(|output| output.status.success())
        .map(|output| String::from_utf8_lossy(&output.stdout).trim().to_owned())
    };
    match (resolve(git_ref), resolve("HEAD")) {
        (Some(base), Some(head)) => base == head,
        _ => false,
    }
}

/// `rev^{commit}`: git resolves it only when `rev` names a commit.
fn peel_to_commit(rev: &str) -> String {
    format!("{rev}^{{commit}}")
}

fn run_git(dir: &Path, args: &[&str]) -> Result<std::process::Output, BaseBaselineError> {
    crate::git_env::git_command()
        .args(args)
        .current_dir(dir)
        .output()
        .map_err(|error| BaseBaselineError::GitMissing(error.to_string()))
}

#[cfg(test)]
mod tests {
    use super::*;
    use serde_json::json;

    fn keys(growth: &BaselineGrowth) -> Vec<(String, Vec<String>)> {
        growth
            .categories
            .iter()
            .map(|grown| (grown.category.clone(), grown.keys.clone()))
            .collect()
    }

    #[test]
    fn dead_code_growth_is_each_new_key_per_category() {
        let base = json!({
            "kind": "dead-code",
            "unused_files": ["src/a.ts"],
            "boundary_violations": ["src/ui/a.ts->src/core/x.ts"],
        });
        let head = json!({
            "kind": "dead-code",
            "unused_files": ["src/a.ts", "src/b.ts"],
            "boundary_violations": [
                "src/ui/a.ts->src/core/x.ts",
                "src/ui/c.ts->src/core/z.ts"
            ],
            "unused_exports": ["src/c.ts:x"],
        });
        let growth = baseline_growth(BaselineKind::DeadCode, &base, &head);
        assert_eq!(
            keys(&growth),
            vec![
                (
                    "boundary_violations".to_owned(),
                    vec!["src/ui/c.ts->src/core/z.ts".to_owned()]
                ),
                ("unused_exports".to_owned(), vec!["src/c.ts:x".to_owned()]),
                ("unused_files".to_owned(), vec!["src/b.ts".to_owned()]),
            ]
        );
        assert_eq!(growth.added_entries(), 3);
    }

    #[test]
    fn dead_code_growth_with_canonical_keys_counts_occurrences() {
        let base = json!({
            "identity": "dc1",
            "unused_class_members": ["unused-class-member:src/a.ts:A:value"],
        });
        let head = json!({
            "identity": "dc1",
            "unused_class_members": [
                "unused-class-member:src/a.ts:A:value",
                "unused-class-member:src/a.ts:A:value"
            ],
        });

        let growth = baseline_growth(BaselineKind::DeadCode, &base, &head);

        assert_eq!(
            keys(&growth),
            vec![(
                "unused_class_members".to_owned(),
                vec!["unused-class-member:src/a.ts:A:value".to_owned()]
            )]
        );
        assert!(baseline_growth(BaselineKind::DeadCode, &head, &base).is_empty());
    }

    #[test]
    fn an_upgrade_that_swaps_a_finding_is_growth() {
        let base = json!({
            "unused_exports": ["src/a.ts:helperA", "src/b.ts:helperB"],
        });
        let head = json!({
            "identity": "dc1",
            "unused_exports": [
                "unused-export:src/b.ts:helperB",
                "unused-export:src/c.ts:helperC"
            ],
        });

        let growth = baseline_growth(BaselineKind::DeadCode, &base, &head);

        assert_eq!(
            keys(&growth),
            vec![(
                "unused_exports".to_owned(),
                vec!["unused-export:src/c.ts:helperC".to_owned()]
            )]
        );
    }

    #[test]
    fn a_clean_upgrade_is_not_growth() {
        let base = json!({
            "unused_files": ["src/old.ts"],
            "unused_exports": ["src/a.ts:helperA"],
            "unused_dependencies": ["packages/app/package.json:lodash"],
            "unused_class_members": ["src/s.ts:Service.run"],
            "unlisted_dependencies": ["chalk"],
            "duplicate_exports": ["Config|src/b.ts|src/a.ts"],
            "circular_dependencies": ["src/a.ts->src/b.ts"],
            "boundary_violations": ["src/ui/a.ts->src/db/q.ts"],
            "stale_suppressions": ["stale-suppression:src/f.ts:3"],
        });
        let head = json!({
            "identity": "dc1",
            "unused_files": ["unused-file:src/old.ts"],
            "unused_exports": ["unused-export:src/a.ts:helperA"],
            "unused_dependencies": ["unused-dependency:packages/app/package.json:lodash"],
            "unused_class_members": ["unused-class-member:src/s.ts:Service:run"],
            "unlisted_dependencies": ["unlisted-dependency:chalk"],
            "duplicate_exports": ["duplicate-export:Config:src/a.ts|src/b.ts"],
            "circular_dependencies": ["circular-dependency:src/a.ts|src/b.ts"],
            "boundary_violations": ["boundary-violation:src/ui/a.ts:src/db/q.ts"],
            "stale_suppressions": ["stale-suppression:src/f.ts:comment:unused-export:line"],
        });

        assert!(baseline_growth(BaselineKind::DeadCode, &base, &head).is_empty());
    }

    #[test]
    fn a_legacy_key_without_a_translation_falls_back_to_the_entry_count() {
        let base = json!({ "unused_dependencies": ["lodash"] });
        let head = json!({
            "identity": "dc1",
            "unused_dependencies": [
                "unused-dependency:package.json:lodash",
                "unused-dependency:package.json:chalk"
            ],
        });

        assert_eq!(
            baseline_growth(BaselineKind::DeadCode, &base, &head).added_entries(),
            1
        );
    }

    #[test]
    fn a_rewritten_legacy_baseline_is_compared_by_entry_counts() {
        let base = json!({ "unused_files": ["src/a.ts", "src/b.ts"] });
        let same = json!({
            "identity": "dc1",
            "unused_files": ["unused-file:src/a.ts", "unused-file:src/b.ts"],
        });
        let grown = json!({
            "identity": "dc1",
            "unused_files": [
                "unused-file:src/a.ts",
                "unused-file:src/b.ts",
                "unused-file:src/c.ts"
            ],
        });

        assert!(baseline_growth(BaselineKind::DeadCode, &base, &same).is_empty());
        assert_eq!(
            baseline_growth(BaselineKind::DeadCode, &base, &grown).added_entries(),
            1
        );
    }

    #[test]
    fn a_removed_or_kept_key_is_not_growth() {
        let base = json!({ "unused_files": ["src/a.ts", "src/b.ts"] });
        let head = json!({ "unused_files": ["src/b.ts"] });
        assert!(baseline_growth(BaselineKind::DeadCode, &base, &head).is_empty());
    }

    #[test]
    fn dupes_growth_counts_one_key_per_clone_group() {
        let base = json!({
            "clone_groups": ["a.ts:1-9|b.ts:1-9"],
            "clone_fingerprints": ["f1"],
            "normalized_clone_fingerprints": ["n1"],
        });
        let head = json!({
            "clone_groups": ["a.ts:1-9|b.ts:1-9", "c.ts:1-9|d.ts:1-9"],
            "clone_fingerprints": ["f1", "f2"],
            "normalized_clone_fingerprints": ["n1", "n2"],
        });
        let growth = baseline_growth(BaselineKind::Dupes, &base, &head);
        assert_eq!(
            keys(&growth),
            vec![(DUPES_FINGERPRINTS.to_owned(), vec!["n2".to_owned()])]
        );
    }

    #[test]
    fn a_legacy_dupes_base_is_compared_by_clone_group_keys() {
        let base = json!({ "clone_groups": ["a.ts:1-9|b.ts:1-9"] });
        let head = json!({
            "clone_groups": ["a.ts:1-9|b.ts:1-9"],
            "normalized_clone_fingerprints": ["n1"],
        });
        assert!(baseline_growth(BaselineKind::Dupes, &base, &head).is_empty());
    }

    #[test]
    fn a_higher_health_count_is_growth() {
        let base = json!({
            "finding_counts": { "src/a.ts": { "cyclomatic": { "count": 1 } } },
        });
        let head = json!({
            "finding_counts": {
                "src/a.ts": { "cyclomatic": { "count": 2 } },
                "src/b.ts": { "cognitive": { "count": 1 } },
            },
            "target_keys": ["src/b.ts:split"],
        });
        let growth = baseline_growth(BaselineKind::Health, &base, &head);
        assert_eq!(
            keys(&growth),
            vec![
                (
                    HEALTH_FILE_COUNTS.to_owned(),
                    vec![
                        "src/a.ts cyclomatic (count 1 -> 2)".to_owned(),
                        "src/b.ts cognitive (count 1)".to_owned(),
                    ]
                ),
                ("target_keys".to_owned(), vec!["src/b.ts:split".to_owned()]),
            ]
        );
        assert_eq!(growth.added_entries(), 3);
    }

    #[test]
    fn identity_counts_replace_file_counts_when_both_files_have_them() {
        let base = json!({
            "finding_counts": { "src/a.ts": { "cyclomatic": { "count": 1 } } },
            "identity_finding_counts": { "src/a.ts\u{0}old": { "cyclomatic": { "count": 1 } } },
        });
        let head = json!({
            "finding_counts": { "src/a.ts": { "cyclomatic": { "count": 1 } } },
            "identity_finding_counts": { "src/a.ts\u{0}new": { "cyclomatic": { "count": 1 } } },
        });
        let growth = baseline_growth(BaselineKind::Health, &base, &head);
        assert_eq!(
            keys(&growth),
            vec![(
                HEALTH_IDENTITY_COUNTS.to_owned(),
                vec!["src/a.ts:new cyclomatic (count 1)".to_owned()]
            )]
        );
    }
}