/** * Log-level types and helpers used by Effect logging. The module defines all * accepted log levels, the concrete emitted severities, the ordered level list, * equality and ordering instances, threshold comparison helpers, and an effect * for checking whether a level is enabled by the current logging settings. * * @since 2.0.0 */ import type * as Effect from "./Effect.ts" import * as Equ from "./Equivalence.ts" import * as core from "./internal/core.ts" import * as effect from "./internal/effect.ts" import * as Ord from "./Order.ts" import * as References from "./References.ts" /** * Represents every level used by Effect logging, including concrete message * severities and the `All` and `None` sentinel levels. * * **When to use** * * Use to type values that may be either concrete log message severities or * logging configuration sentinels. * * **Details** * * The levels are ordered from most severe to least severe: * - `All` - Special level that allows all messages * - `Fatal` - System is unusable, immediate attention required * - `Error` - Error conditions that should be investigated * - `Warn` - Warning conditions that may indicate problems * - `Info` - Informational messages about normal operation * - `Debug` - Debug information useful during development * - `Trace` - Very detailed trace information * - `None` - Special level that suppresses all messages * * **Example** (Using log levels) * * ```ts * import { Effect } from "effect" * * // Using log levels with Effect logging * const program = Effect.gen(function*() { * yield* Effect.logFatal("System failure") * yield* Effect.logError("Database error") * yield* Effect.logWarning("High memory usage") * yield* Effect.logInfo("User logged in") * yield* Effect.logDebug("Processing request") * yield* Effect.logTrace("Variable state") * }) * * // Type-safe log level variables * const errorLevel = "Error" // LogLevel * const debugLevel = "Debug" // LogLevel * ``` * * @category models * @since 2.0.0 */ export type LogLevel = "All" | "Fatal" | "Error" | "Warn" | "Info" | "Debug" | "Trace" | "None" /** * Log levels that represent actual message severities, excluding the `All` and * `None` sentinel levels. * * **When to use** * * Use when typing emitted log message severities, such as explicit log calls, * current log level references, or error-report severity annotations, where * `All` and `None` are not valid values. * * @see {@link LogLevel} for the wider log-level type that also accepts the * `All` and `None` sentinel levels * @see {@link values} for the runtime list of all accepted `LogLevel` values, * including sentinels * * @category models * @since 4.0.0 */ export type Severity = "Fatal" | "Error" | "Warn" | "Info" | "Debug" | "Trace" /** * Returns all `LogLevel` values in order from `All` through the concrete severities to * `None`. * * **When to use** * * Use to enumerate or validate all accepted `LogLevel` string values, including * the `All` and `None` sentinel levels. * * **Details** * * The array order matches the module severity order: `All`, concrete * severities from `Fatal` to `Trace`, then `None`. * * **Gotchas** * * This list includes `All` and `None`, so it is not limited to concrete emitted * severities. * * @see {@link Severity} for the concrete message severity type that excludes `All` and `None` * @see {@link Order} for comparing these levels by severity order * * @category models * @since 4.0.0 */ export const values: ReadonlyArray = ["All", "Fatal", "Error", "Warn", "Info", "Debug", "Trace", "None"] /** * Order instance for `LogLevel` that defines the severity ordering. * * **When to use** * * Use to sort or compare log levels according to Effect's severity order. * * **Details** * * This order treats "All" as the least restrictive level and "None" as the most restrictive, * with Fatal being the most severe actual log level. * * **Example** (Ordering log levels) * * ```ts * import { LogLevel } from "effect" * * // Compare log levels using Order * console.log(LogLevel.Order("Error", "Info")) // 1 (Error > Info) * console.log(LogLevel.Order("Debug", "Error")) // -1 (Debug < Error) * console.log(LogLevel.Order("Info", "Info")) // 0 (Info == Info) * ``` * * @category ordering * @since 2.0.0 */ export const Order: Ord.Order = effect.LogLevelOrder /** * Equivalence instance for log levels using strict equality (`===`). * * **When to use** * * Use to compare two `LogLevel` values when only the exact same level should * match. * * **Details** * * Each log level string, including `All` and `None`, only matches itself. * * **Example** (Comparing log levels) * * ```ts * import { LogLevel } from "effect" * * console.log(LogLevel.Equivalence("Error", "Error")) // true * console.log(LogLevel.Equivalence("Error", "Info")) // false * ``` * * @see {@link Order} for severity ordering rather than exact level equality * @see {@link isGreaterThanOrEqualTo} for minimum-threshold checks * * @category instances * @since 4.0.0 */ export const Equivalence: Equ.Equivalence = Equ.strictEqual() /** * Returns the ordinal value of the log level. * * **When to use** * * Use to project a `LogLevel` into the numeric sort key used by * `LogLevel.Order` when custom ordering code or an integration needs a number * instead of an `Order` comparison. * * **Details** * * The mapping is `All` to `Number.MIN_SAFE_INTEGER`, `Trace` to `0`, `Debug` to * `10000`, `Info` to `20000`, `Warn` to `30000`, `Error` to `40000`, `Fatal` to * `50000`, and `None` to `Number.MAX_SAFE_INTEGER`. * * **Gotchas** * * These ordinals are internal sort keys; do not treat them as external severity * numbers. * * @see {@link Order} for comparing log levels without exposing numeric keys * @see {@link isGreaterThanOrEqualTo} for minimum-threshold filtering * * @category ordering * @since 4.0.0 */ export const getOrdinal = (self: LogLevel): number => effect.logLevelToOrder(self) /** * Determines if the first log level is more severe than the second. * * **When to use** * * Use to check whether one log level is strictly more severe than another. * * **Details** * * Returns `true` if `self` represents a more severe level than `that`. * * **Example** (Checking higher severity) * * ```ts * import { LogLevel } from "effect" * * // Check if Error is more severe than Info * console.log(LogLevel.isGreaterThan("Error", "Info")) // true * console.log(LogLevel.isGreaterThan("Debug", "Error")) // false * * // Use with filtering * const isFatal = LogLevel.isGreaterThan("Fatal", "Warn") * const isError = LogLevel.isGreaterThan("Error", "Warn") * const isDebug = LogLevel.isGreaterThan("Debug", "Warn") * console.log(isFatal) // true * console.log(isError) // true * console.log(isDebug) // false * * // Curried usage * const isMoreSevereThanInfo = LogLevel.isGreaterThan("Info") * console.log(isMoreSevereThanInfo("Error")) // true * console.log(isMoreSevereThanInfo("Debug")) // false * ``` * * @category ordering * @since 4.0.0 */ export const isGreaterThan: { (that: LogLevel): (self: LogLevel) => boolean (self: LogLevel, that: LogLevel): boolean } = effect.isLogLevelGreaterThan /** * Determines if the first log level is more severe than or equal to the second. * * **When to use** * * Use to implement minimum log-level filtering by checking whether a message * level meets a threshold. * * **Details** * * Returns `true` if `self` represents a level that is more severe than or equal to `that`. * * **Example** (Filtering by minimum log level) * * ```ts * import { Logger, LogLevel } from "effect" * * // Check if level meets minimum threshold * console.log(LogLevel.isGreaterThanOrEqualTo("Error", "Error")) // true * console.log(LogLevel.isGreaterThanOrEqualTo("Error", "Info")) // true * console.log(LogLevel.isGreaterThanOrEqualTo("Debug", "Info")) // false * * // Create a logger that only logs Info and above * const infoLogger = Logger.make((options) => { * if (LogLevel.isGreaterThanOrEqualTo(options.logLevel, "Info")) { * console.log(`[${options.logLevel}] ${options.message}`) * } * }) * * // Production logger - only Error and Fatal * const productionLogger = Logger.make((options) => { * if (LogLevel.isGreaterThanOrEqualTo(options.logLevel, "Error")) { * console.error( * `${options.date.toISOString()} [${options.logLevel}] ${options.message}` * ) * } * }) * * // Curried usage for filtering * const isInfoOrAbove = LogLevel.isGreaterThanOrEqualTo("Info") * const shouldLog = isInfoOrAbove("Error") // true * ``` * * @category ordering * @since 4.0.0 */ export const isGreaterThanOrEqualTo: { (that: LogLevel): (self: LogLevel) => boolean (self: LogLevel, that: LogLevel): boolean } = Ord.isGreaterThanOrEqualTo(Order) /** * Determines if the first log level is less severe than the second. * * **When to use** * * Use to check whether one log level is strictly less severe than another. * * **Details** * * Returns `true` if `self` represents a less severe level than `that`. * * **Example** (Checking lower severity) * * ```ts * import { LogLevel } from "effect" * * // Check if Debug is less severe than Info * console.log(LogLevel.isLessThan("Debug", "Info")) // true * console.log(LogLevel.isLessThan("Error", "Info")) // false * * // Filter out verbose logs * const isFatalVerbose = LogLevel.isLessThan("Fatal", "Info") * const isErrorVerbose = LogLevel.isLessThan("Error", "Info") * const isTraceVerbose = LogLevel.isLessThan("Trace", "Info") * console.log(isFatalVerbose) // false (Fatal is not verbose) * console.log(isErrorVerbose) // false (Error is not verbose) * console.log(isTraceVerbose) // true (Trace is verbose) * * // Curried usage * const isLessSevereThanError = LogLevel.isLessThan("Error") * console.log(isLessSevereThanError("Info")) // true * console.log(isLessSevereThanError("Fatal")) // false * ``` * * @category ordering * @since 4.0.0 */ export const isLessThan: { (that: LogLevel): (self: LogLevel) => boolean (self: LogLevel, that: LogLevel): boolean } = Ord.isLessThan(Order) /** * Determines if the first log level is less severe than or equal to the second. * * **When to use** * * Use to implement maximum log-level filtering by checking whether a level is * at or below a threshold. * * **Details** * * Returns `true` if `self` represents a level that is less severe than or equal to `that`. * * **Example** (Filtering by maximum log level) * * ```ts * import { Logger, LogLevel } from "effect" * * // Check if level is at or below threshold * console.log(LogLevel.isLessThanOrEqualTo("Info", "Info")) // true * console.log(LogLevel.isLessThanOrEqualTo("Debug", "Info")) // true * console.log(LogLevel.isLessThanOrEqualTo("Error", "Info")) // false * * // Create a logger that suppresses verbose logs * const quietLogger = Logger.make((options) => { * if (LogLevel.isLessThanOrEqualTo(options.logLevel, "Info")) { * console.log(`[${options.logLevel}] ${options.message}`) * } * }) * * // Development logger - suppress trace logs * const devLogger = Logger.make((options) => { * if (LogLevel.isLessThanOrEqualTo(options.logLevel, "Debug")) { * console.log(`[${options.logLevel}] ${options.message}`) * } * }) * * // Curried usage for filtering * const isInfoOrBelow = LogLevel.isLessThanOrEqualTo("Info") * const shouldLog = isInfoOrBelow("Debug") // true * ``` * * @category ordering * @since 4.0.0 */ export const isLessThanOrEqualTo: { (that: LogLevel): (self: LogLevel) => boolean (self: LogLevel, that: LogLevel): boolean } = Ord.isLessThanOrEqualTo(Order) /** * Checks whether a given log level is enabled for the current fiber. * * **When to use** * * Use to check whether a log level would be emitted under the current fiber's * minimum log level. * * **Details** * * A log level is enabled when it is greater than or equal to * `References.MinimumLogLevel`. * * **Example** (Checking current fiber log level) * * ```ts * import { Effect, LogLevel, References } from "effect" * * const program = Effect.gen(function*() { * const debugEnabled = yield* LogLevel.isEnabled("Debug") * const errorEnabled = yield* LogLevel.isEnabled("Error") * * console.log({ debugEnabled, errorEnabled }) * }) * * const warnOnly = program.pipe( * Effect.provideService(References.MinimumLogLevel, "Warn") * ) * ``` * * @category filtering * @since 4.0.0 */ export const isEnabled = (self: LogLevel): Effect.Effect => core.withFiber((fiber) => effect.succeed(!isGreaterThan(fiber.getRef(References.MinimumLogLevel), self)))