ic_query/nns/neuron/report/distribution/

mod.rs

//! Module: nns::neuron::report::distribution
//!
//! Responsibility: validate and aggregate complete caller-retained public NNS neuron collections.
//! Does not own: collection transport, persistence, cache policy, or process output.
//! Boundary: projects complete public `NeuronInfo` evidence into portable distributions.

use super::{
    NnsNeuronCollectionState, collection::validate_collection_state, model::NnsNeuronRow,
    source::validate_neuron_rows,
};
use crate::{
    nns::{
        MAINNET_GOVERNANCE_CANISTER_ID,
        governance::{
            NnsGovernanceCollectionStatus, NnsGovernanceSourceProvenance,
            validate_governance_report_source, validate_governance_time_interval,
        },
    },
    subnet_catalog::MAINNET_NETWORK,
};
use serde::{Deserialize, Serialize};
use std::collections::BTreeMap;
use thiserror::Error as ThisError;

/// Version of the portable NNS public-neuron distribution report schema.
pub const NNS_NEURON_DISTRIBUTION_REPORT_SCHEMA_VERSION: u32 = 1;

///
/// NnsNeuronStateDistribution
///
/// Public neuron count and effective stake for one raw Governance state code.
///

#[derive(Clone, Debug, Deserialize, Eq, PartialEq, Serialize)]
#[serde(deny_unknown_fields)]
pub struct NnsNeuronStateDistribution {
    /// Raw native Governance state code.
    pub state: i32,
    /// Number of collected neurons with this state code.
    pub neuron_count: u64,
    /// Sum of public effective stake for neurons with this state code.
    pub effective_stake_e8s: u64,
}

///
/// NnsNeuronVisibilityDistribution
///
/// Public neuron count and effective stake for one optional Governance visibility code.
///

#[derive(Clone, Debug, Deserialize, Eq, PartialEq, Serialize)]
#[serde(deny_unknown_fields)]
pub struct NnsNeuronVisibilityDistribution {
    /// Raw optional native Governance visibility code.
    pub visibility: Option<i32>,
    /// Number of collected neurons with this visibility code or absence.
    pub neuron_count: u64,
    /// Sum of public effective stake for neurons with this visibility code or absence.
    pub effective_stake_e8s: u64,
}

///
/// NnsNeuronTypeDistribution
///
/// Public neuron count and effective stake for one optional Governance neuron-type code.
///

#[derive(Clone, Debug, Deserialize, Eq, PartialEq, Serialize)]
#[serde(deny_unknown_fields)]
pub struct NnsNeuronTypeDistribution {
    /// Raw optional native Governance neuron-type code.
    pub neuron_type: Option<i32>,
    /// Number of collected neurons with this type code or absence.
    pub neuron_count: u64,
    /// Sum of public effective stake for neurons with this type code or absence.
    pub effective_stake_e8s: u64,
}

///
/// NnsNeuronDistributionReport
///
/// Deterministic local distribution over one complete public NNS neuron collection.
///

#[derive(Clone, Debug, Deserialize, Eq, PartialEq, Serialize)]
pub struct NnsNeuronDistributionReport {
    /// Report schema version.
    pub schema_version: u32,
    /// Network identity retained by the complete collection.
    pub network: String,
    /// Fixed NNS Governance canister principal retained by the collection.
    pub governance_canister_id: String,
    /// Concrete transport and collector provenance retained by the collection.
    pub source: NnsGovernanceSourceProvenance,
    /// Caller-supplied time attached to collection creation.
    pub collection_started_at: String,
    /// Caller-supplied time attached to the final admitted page.
    pub collection_updated_at: String,
    /// Number of admitted source pages in the complete collection.
    pub collection_page_count: u32,
    /// Number of admitted public neuron rows in the complete collection.
    pub collected_neuron_count: u64,
    /// Whether the sequential collection is guaranteed to represent one point in time.
    pub point_in_time_guaranteed: bool,
    /// Earliest raw Governance retrieval timestamp among collected rows.
    pub earliest_retrieved_at_timestamp_seconds: Option<u64>,
    /// Latest raw Governance retrieval timestamp among collected rows.
    pub latest_retrieved_at_timestamp_seconds: Option<u64>,
    /// Sum of public effective stake across every collected neuron.
    pub total_effective_stake_e8s: u64,
    /// Rows carrying a public staked-maturity value.
    pub reported_staked_maturity_neuron_count: u64,
    /// Rows without a public staked-maturity value.
    pub unreported_staked_maturity_neuron_count: u64,
    /// Sum of reported staked maturity values only.
    pub total_reported_staked_maturity_e8s_equivalent: u64,
    /// Rows carrying current deciding voting power.
    pub reported_deciding_voting_power_neuron_count: u64,
    /// Rows without current deciding voting power.
    pub unreported_deciding_voting_power_neuron_count: u64,
    /// Sum of reported current deciding voting power values only.
    pub total_reported_deciding_voting_power: u64,
    /// Rows carrying current potential voting power.
    pub reported_potential_voting_power_neuron_count: u64,
    /// Rows without current potential voting power.
    pub unreported_potential_voting_power_neuron_count: u64,
    /// Sum of reported current potential voting power values only.
    pub total_reported_potential_voting_power: u64,
    /// Rows carrying registered known-neuron metadata.
    pub known_neuron_metadata_count: u64,
    /// Rows carrying a public Neurons' Fund join timestamp.
    pub neurons_fund_join_timestamp_present_count: u64,
    /// Canonically raw-code-ordered distribution by neuron state.
    pub state_distribution: Vec<NnsNeuronStateDistribution>,
    /// Canonically optional-raw-code-ordered distribution by visibility.
    pub visibility_distribution: Vec<NnsNeuronVisibilityDistribution>,
    /// Canonically optional-raw-code-ordered distribution by neuron type.
    pub neuron_type_distribution: Vec<NnsNeuronTypeDistribution>,
}

///
/// NnsNeuronDistributionValidationError
///
/// Pure validation failure for an untrusted serialized or in-memory distribution report.
///

#[derive(Debug, Eq, PartialEq, ThisError)]
#[error("invalid NNS neuron distribution report: {reason}")]
pub struct NnsNeuronDistributionValidationError {
    /// Deterministic invariant failure.
    pub reason: String,
}

///
/// NnsNeuronDistributionError
///
/// Deterministic validation or accounting failure from local neuron distribution projection.
///

#[derive(Debug, ThisError)]
pub enum NnsNeuronDistributionError {
    /// The supplied collection state failed its shared continuation invariants.
    #[error("invalid NNS neuron collection state for distribution projection: {reason}")]
    InvalidCollectionState {
        /// Deterministic collection invariant failure.
        reason: String,
    },

    /// The collection stopped without observing Governance API exhaustion.
    #[error("NNS neuron distribution requires a complete collection; state is {status}")]
    CollectionNotComplete {
        /// Current lifecycle of the otherwise valid collection state.
        status: NnsGovernanceCollectionStatus,
    },

    /// The supplied rows do not match the collection's admitted-row accounting.
    #[error(
        "NNS neuron distribution received {actual} rows; complete collection accounts for {expected}"
    )]
    NeuronCountMismatch {
        /// Neuron rows accounted for by the collection state.
        expected: u64,
        /// Neuron rows supplied to the builder.
        actual: u64,
    },

    /// Supplied rows failed the shared public-neuron response contract.
    #[error("invalid NNS neuron rows for distribution projection: {reason}")]
    InvalidNeuronRows {
        /// Deterministic row invariant failure.
        reason: String,
    },

    /// A count conversion, increment, or numeric sum exceeded `u64`.
    #[error("NNS neuron distribution accounting overflow while updating {field}")]
    AccountingOverflow {
        /// Count or sum that exceeded its representation.
        field: &'static str,
    },
}

/// Build one deterministic distribution from a complete caller-retained public-neuron collection.
pub fn build_nns_neuron_distribution_report(
    collection: &NnsNeuronCollectionState,
    neurons: &[NnsNeuronRow],
) -> Result<NnsNeuronDistributionReport, NnsNeuronDistributionError> {
    validate_collection_state(collection).map_err(|error| {
        NnsNeuronDistributionError::InvalidCollectionState {
            reason: error.to_string(),
        }
    })?;
    if !collection.is_complete() {
        return Err(NnsNeuronDistributionError::CollectionNotComplete {
            status: collection.status(),
        });
    }
    let source = collection.source().cloned().ok_or_else(|| {
        NnsNeuronDistributionError::InvalidCollectionState {
            reason: "complete collection has no concrete source provenance".to_string(),
        }
    })?;

    let expected = collection.neurons_fetched();
    let actual = u64::try_from(neurons.len()).map_err(|_| {
        NnsNeuronDistributionError::AccountingOverflow {
            field: "supplied_neuron_count",
        }
    })?;
    if actual != expected {
        return Err(NnsNeuronDistributionError::NeuronCountMismatch { expected, actual });
    }
    validate_neuron_rows(neurons).map_err(|error| {
        NnsNeuronDistributionError::InvalidNeuronRows {
            reason: error.to_string(),
        }
    })?;

    let mut distribution = DistributionAccumulator::default();
    for neuron in neurons {
        distribution.observe(neuron)?;
    }
    Ok(distribution.into_report(collection, expected, source))
}

/// Validate every distribution-report invariant available without source rows or live host calls.
pub fn validate_nns_neuron_distribution_report(
    report: &NnsNeuronDistributionReport,
) -> Result<(), NnsNeuronDistributionValidationError> {
    validate_distribution_header(report)?;
    validate_distribution_summary(report)?;
    validate_distribution(
        report,
        "state",
        "raw-code",
        report
            .state_distribution
            .iter()
            .map(|row| (row.state, row.neuron_count, row.effective_stake_e8s)),
    )?;
    validate_distribution(
        report,
        "visibility",
        "optional-raw-code",
        report
            .visibility_distribution
            .iter()
            .map(|row| (row.visibility, row.neuron_count, row.effective_stake_e8s)),
    )?;
    validate_distribution(
        report,
        "neuron-type",
        "optional-raw-code",
        report
            .neuron_type_distribution
            .iter()
            .map(|row| (row.neuron_type, row.neuron_count, row.effective_stake_e8s)),
    )
}

fn validate_distribution_header(
    report: &NnsNeuronDistributionReport,
) -> Result<(), NnsNeuronDistributionValidationError> {
    if report.schema_version != NNS_NEURON_DISTRIBUTION_REPORT_SCHEMA_VERSION {
        return Err(invalid_validation(format!(
            "schema version {} does not equal {}",
            report.schema_version, NNS_NEURON_DISTRIBUTION_REPORT_SCHEMA_VERSION
        )));
    }
    if report.network != MAINNET_NETWORK {
        return Err(invalid_validation(format!(
            "network is {}, expected {MAINNET_NETWORK}",
            report.network
        )));
    }
    if report.governance_canister_id != MAINNET_GOVERNANCE_CANISTER_ID {
        return Err(invalid_validation(format!(
            "governance_canister_id is {}, expected {MAINNET_GOVERNANCE_CANISTER_ID}",
            report.governance_canister_id
        )));
    }
    if report.collection_page_count == 0 {
        return Err(invalid_validation(
            "complete distribution report must retain at least one collection page",
        ));
    }
    let minimum_neuron_count = u64::from(report.collection_page_count - 1);
    if report.collected_neuron_count < minimum_neuron_count {
        return Err(invalid_validation(format!(
            "collection_page_count {} requires at least {minimum_neuron_count} collected neurons, found {}",
            report.collection_page_count, report.collected_neuron_count
        )));
    }
    validate_governance_time_interval(
        "collection_started_at",
        &report.collection_started_at,
        "collection_updated_at",
        &report.collection_updated_at,
    )
    .map_err(|error| invalid_validation(error.to_string()))?;
    if report.point_in_time_guaranteed {
        return Err(invalid_validation(
            "sequential public-neuron collection cannot claim a point-in-time snapshot",
        ));
    }
    validate_governance_report_source(&report.network, &report.source).map_err(|error| {
        let context = match &report.source {
            NnsGovernanceSourceProvenance::ReplicaQuery { .. } => "source",
            NnsGovernanceSourceProvenance::ReplicatedInterCanisterCall { .. } => "provenance",
        };
        invalid_validation(format!("invalid collection {context}: {error}"))
    })?;
    validate_retrieval_range(report)
}

fn validate_retrieval_range(
    report: &NnsNeuronDistributionReport,
) -> Result<(), NnsNeuronDistributionValidationError> {
    match (
        report.earliest_retrieved_at_timestamp_seconds,
        report.latest_retrieved_at_timestamp_seconds,
    ) {
        (None, None) if report.collected_neuron_count == 0 => Ok(()),
        (Some(earliest), Some(latest))
            if report.collected_neuron_count > 0 && earliest <= latest =>
        {
            Ok(())
        }
        _ => Err(invalid_validation(
            "retrieval timestamp range disagrees with collected_neuron_count or is reversed",
        )),
    }
}

fn validate_distribution_summary(
    report: &NnsNeuronDistributionReport,
) -> Result<(), NnsNeuronDistributionValidationError> {
    validate_optional_summary(
        report.reported_staked_maturity_neuron_count,
        report.unreported_staked_maturity_neuron_count,
        report.total_reported_staked_maturity_e8s_equivalent,
        report.collected_neuron_count,
        "staked maturity",
    )?;
    validate_optional_summary(
        report.reported_deciding_voting_power_neuron_count,
        report.unreported_deciding_voting_power_neuron_count,
        report.total_reported_deciding_voting_power,
        report.collected_neuron_count,
        "deciding voting power",
    )?;
    validate_optional_summary(
        report.reported_potential_voting_power_neuron_count,
        report.unreported_potential_voting_power_neuron_count,
        report.total_reported_potential_voting_power,
        report.collected_neuron_count,
        "potential voting power",
    )?;
    for (field, count) in [
        (
            "known_neuron_metadata_count",
            report.known_neuron_metadata_count,
        ),
        (
            "neurons_fund_join_timestamp_present_count",
            report.neurons_fund_join_timestamp_present_count,
        ),
    ] {
        if count > report.collected_neuron_count {
            return Err(invalid_validation(format!(
                "{field} {count} exceeds collected_neuron_count {}",
                report.collected_neuron_count
            )));
        }
    }
    Ok(())
}

fn validate_optional_summary(
    reported: u64,
    unreported: u64,
    total: u64,
    collected: u64,
    field: &'static str,
) -> Result<(), NnsNeuronDistributionValidationError> {
    let accounted = reported
        .checked_add(unreported)
        .ok_or_else(|| invalid_validation(format!("{field} coverage count overflow")))?;
    if accounted != collected {
        return Err(invalid_validation(format!(
            "{field} coverage accounts for {accounted} neurons, expected {collected}"
        )));
    }
    if reported == 0 && total != 0 {
        return Err(invalid_validation(format!(
            "{field} total must be zero when no rows report the field"
        )));
    }
    Ok(())
}

fn validate_distribution<Key: Copy + Ord + std::fmt::Debug>(
    report: &NnsNeuronDistributionReport,
    dimension: &'static str,
    order: &'static str,
    rows: impl Iterator<Item = (Key, u64, u64)>,
) -> Result<(), NnsNeuronDistributionValidationError> {
    let mut previous = None;
    let mut neuron_count = 0_u64;
    let mut stake = 0_u64;
    for (key, count, effective_stake) in rows {
        if previous.is_some_and(|previous| previous >= key) {
            return Err(invalid_validation(format!(
                "{dimension} distribution is not strictly {order} ordered"
            )));
        }
        neuron_count = add_distribution_count(neuron_count, count, dimension)?;
        stake = stake
            .checked_add(effective_stake)
            .ok_or_else(|| invalid_validation(format!("{dimension} stake total overflow")))?;
        previous = Some(key);
    }
    validate_distribution_totals(report, neuron_count, stake, dimension)
}

fn add_distribution_count(
    total: u64,
    count: u64,
    dimension: &'static str,
) -> Result<u64, NnsNeuronDistributionValidationError> {
    if count == 0 {
        return Err(invalid_validation(format!(
            "{dimension} distribution row must contain at least one neuron"
        )));
    }
    add_validation_total(total, count, dimension)
}

fn add_validation_total(
    total: u64,
    value: u64,
    field: &'static str,
) -> Result<u64, NnsNeuronDistributionValidationError> {
    total
        .checked_add(value)
        .ok_or_else(|| invalid_validation(format!("{field} total overflow")))
}

fn validate_distribution_totals(
    report: &NnsNeuronDistributionReport,
    neuron_count: u64,
    stake: u64,
    dimension: &'static str,
) -> Result<(), NnsNeuronDistributionValidationError> {
    if neuron_count != report.collected_neuron_count {
        return Err(invalid_validation(format!(
            "{dimension} neuron counts sum to {neuron_count}, expected {}",
            report.collected_neuron_count
        )));
    }
    if stake != report.total_effective_stake_e8s {
        return Err(invalid_validation(format!(
            "{dimension} effective stake sums to {stake}, expected {}",
            report.total_effective_stake_e8s
        )));
    }
    Ok(())
}

fn invalid_validation(reason: impl Into<String>) -> NnsNeuronDistributionValidationError {
    NnsNeuronDistributionValidationError {
        reason: reason.into(),
    }
}

#[derive(Clone, Copy, Default)]
struct DimensionAccumulator {
    neuron_count: u64,
    effective_stake_e8s: u64,
}

impl DimensionAccumulator {
    fn observe(
        &mut self,
        effective_stake_e8s: u64,
        count_field: &'static str,
        stake_field: &'static str,
    ) -> Result<(), NnsNeuronDistributionError> {
        increment(&mut self.neuron_count, count_field)?;
        add(
            &mut self.effective_stake_e8s,
            effective_stake_e8s,
            stake_field,
        )
    }
}

#[derive(Default)]
struct OptionalValueAccumulator {
    reported_neuron_count: u64,
    unreported_neuron_count: u64,
    total_reported_value: u64,
}

impl OptionalValueAccumulator {
    fn observe(
        &mut self,
        value: Option<u64>,
        reported_field: &'static str,
        unreported_field: &'static str,
        total_field: &'static str,
    ) -> Result<(), NnsNeuronDistributionError> {
        if let Some(value) = value {
            increment(&mut self.reported_neuron_count, reported_field)?;
            add(&mut self.total_reported_value, value, total_field)
        } else {
            increment(&mut self.unreported_neuron_count, unreported_field)
        }
    }
}

#[derive(Default)]
struct DistributionAccumulator {
    states: BTreeMap<i32, DimensionAccumulator>,
    visibilities: BTreeMap<Option<i32>, DimensionAccumulator>,
    neuron_types: BTreeMap<Option<i32>, DimensionAccumulator>,
    total_effective_stake_e8s: u64,
    staked_maturity: OptionalValueAccumulator,
    deciding_voting_power: OptionalValueAccumulator,
    potential_voting_power: OptionalValueAccumulator,
    known_neuron_metadata_count: u64,
    neurons_fund_join_timestamp_present_count: u64,
    earliest_retrieved_at_timestamp_seconds: Option<u64>,
    latest_retrieved_at_timestamp_seconds: Option<u64>,
}

impl DistributionAccumulator {
    fn observe(&mut self, neuron: &NnsNeuronRow) -> Result<(), NnsNeuronDistributionError> {
        add(
            &mut self.total_effective_stake_e8s,
            neuron.stake_e8s,
            "total_effective_stake_e8s",
        )?;
        self.states.entry(neuron.state).or_default().observe(
            neuron.stake_e8s,
            "state_neuron_count",
            "state_effective_stake_e8s",
        )?;
        self.visibilities
            .entry(neuron.visibility)
            .or_default()
            .observe(
                neuron.stake_e8s,
                "visibility_neuron_count",
                "visibility_effective_stake_e8s",
            )?;
        self.neuron_types
            .entry(neuron.neuron_type)
            .or_default()
            .observe(
                neuron.stake_e8s,
                "neuron_type_neuron_count",
                "neuron_type_effective_stake_e8s",
            )?;
        self.staked_maturity.observe(
            neuron.staked_maturity_e8s_equivalent,
            "reported_staked_maturity_neuron_count",
            "unreported_staked_maturity_neuron_count",
            "total_reported_staked_maturity_e8s_equivalent",
        )?;
        self.deciding_voting_power.observe(
            neuron.deciding_voting_power,
            "reported_deciding_voting_power_neuron_count",
            "unreported_deciding_voting_power_neuron_count",
            "total_reported_deciding_voting_power",
        )?;
        self.potential_voting_power.observe(
            neuron.potential_voting_power,
            "reported_potential_voting_power_neuron_count",
            "unreported_potential_voting_power_neuron_count",
            "total_reported_potential_voting_power",
        )?;
        if neuron.known_neuron_data.is_some() {
            increment(
                &mut self.known_neuron_metadata_count,
                "known_neuron_metadata_count",
            )?;
        }
        if neuron.joined_community_fund_timestamp_seconds.is_some() {
            increment(
                &mut self.neurons_fund_join_timestamp_present_count,
                "neurons_fund_join_timestamp_present_count",
            )?;
        }
        let retrieved_at = neuron.retrieved_at_timestamp_seconds;
        self.earliest_retrieved_at_timestamp_seconds = Some(
            self.earliest_retrieved_at_timestamp_seconds
                .map_or(retrieved_at, |earliest| earliest.min(retrieved_at)),
        );
        self.latest_retrieved_at_timestamp_seconds = Some(
            self.latest_retrieved_at_timestamp_seconds
                .map_or(retrieved_at, |latest| latest.max(retrieved_at)),
        );
        Ok(())
    }

    fn into_report(
        self,
        collection: &NnsNeuronCollectionState,
        collected_neuron_count: u64,
        source: NnsGovernanceSourceProvenance,
    ) -> NnsNeuronDistributionReport {
        NnsNeuronDistributionReport {
            schema_version: NNS_NEURON_DISTRIBUTION_REPORT_SCHEMA_VERSION,
            network: collection.network().to_string(),
            governance_canister_id: collection.governance_canister_id().to_string(),
            source,
            collection_started_at: collection.started_at().to_string(),
            collection_updated_at: collection.updated_at().to_string(),
            collection_page_count: collection.pages_fetched(),
            collected_neuron_count,
            point_in_time_guaranteed: false,
            earliest_retrieved_at_timestamp_seconds: self.earliest_retrieved_at_timestamp_seconds,
            latest_retrieved_at_timestamp_seconds: self.latest_retrieved_at_timestamp_seconds,
            total_effective_stake_e8s: self.total_effective_stake_e8s,
            reported_staked_maturity_neuron_count: self.staked_maturity.reported_neuron_count,
            unreported_staked_maturity_neuron_count: self.staked_maturity.unreported_neuron_count,
            total_reported_staked_maturity_e8s_equivalent: self
                .staked_maturity
                .total_reported_value,
            reported_deciding_voting_power_neuron_count: self
                .deciding_voting_power
                .reported_neuron_count,
            unreported_deciding_voting_power_neuron_count: self
                .deciding_voting_power
                .unreported_neuron_count,
            total_reported_deciding_voting_power: self.deciding_voting_power.total_reported_value,
            reported_potential_voting_power_neuron_count: self
                .potential_voting_power
                .reported_neuron_count,
            unreported_potential_voting_power_neuron_count: self
                .potential_voting_power
                .unreported_neuron_count,
            total_reported_potential_voting_power: self.potential_voting_power.total_reported_value,
            known_neuron_metadata_count: self.known_neuron_metadata_count,
            neurons_fund_join_timestamp_present_count: self
                .neurons_fund_join_timestamp_present_count,
            state_distribution: self
                .states
                .into_iter()
                .map(|(state, distribution)| NnsNeuronStateDistribution {
                    state,

                    neuron_count: distribution.neuron_count,
                    effective_stake_e8s: distribution.effective_stake_e8s,
                })
                .collect(),
            visibility_distribution: self
                .visibilities
                .into_iter()
                .map(
                    |(visibility, distribution)| NnsNeuronVisibilityDistribution {
                        visibility,

                        neuron_count: distribution.neuron_count,
                        effective_stake_e8s: distribution.effective_stake_e8s,
                    },
                )
                .collect(),
            neuron_type_distribution: self
                .neuron_types
                .into_iter()
                .map(|(neuron_type, distribution)| NnsNeuronTypeDistribution {
                    neuron_type,

                    neuron_count: distribution.neuron_count,
                    effective_stake_e8s: distribution.effective_stake_e8s,
                })
                .collect(),
        }
    }
}

fn increment(value: &mut u64, field: &'static str) -> Result<(), NnsNeuronDistributionError> {
    *value = value
        .checked_add(1)
        .ok_or(NnsNeuronDistributionError::AccountingOverflow { field })?;
    Ok(())
}

fn add(total: &mut u64, value: u64, field: &'static str) -> Result<(), NnsNeuronDistributionError> {
    *total = total
        .checked_add(value)
        .ok_or(NnsNeuronDistributionError::AccountingOverflow { field })?;
    Ok(())
}

#[cfg(test)]
mod tests;