tracing_subscriber/filter/env/

mod.rs

//! A `Layer` that enables or disables spans and events based on a set of
//! filtering directives.

// these are publicly re-exported, but the compiler doesn't realize
// that for some reason.
#[allow(unreachable_pub)]
pub use self::{builder::Builder, directive::Directive, field::BadName as BadFieldName};
mod builder;
mod directive;
mod field;

use crate::{
    filter::LevelFilter,
    layer::{Context, Layer},
    sync::RwLock,
};
use alloc::{fmt, str::FromStr, vec::Vec};
use core::cell::RefCell;
use directive::ParseError;
use std::{collections::HashMap, env, error::Error};
use thread_local::ThreadLocal;
use tracing_core::{
    callsite,
    field::Field,
    span,
    subscriber::{Interest, Subscriber},
    Metadata,
};

/// A [`Layer`] which filters spans and events based on a set of filter
/// directives.
///
/// `EnvFilter` implements both the [`Layer`](#impl-Layer<S>) and [`Filter`] traits, so it may
/// be used for both [global filtering][global] and [per-layer filtering][plf],
/// respectively. See [the documentation on filtering with `Layer`s][filtering]
/// for details.
///
/// The [`Targets`] type implements a similar form of filtering, but without the
/// ability to dynamically enable events based on the current span context, and
/// without filtering on field values. When these features are not required,
/// [`Targets`] provides a lighter-weight alternative to [`EnvFilter`].
///
/// # Directives
///
/// A filter consists of one or more comma-separated directives which match on [`Span`]s and [`Event`]s.
/// Each directive may have a corresponding maximum verbosity [`level`] which
/// enables (e.g., _selects for_) spans and events that match. Like `log`,
/// `tracing` considers less exclusive levels (like `trace` or `info`) to be more
/// verbose than more exclusive levels (like `error` or `warn`).
///
/// The directive syntax is similar to that of [`env_logger`]'s. At a high level, the syntax for directives
/// consists of several parts:
///
/// ```text
/// target[span{field=value}]=level
/// ```
///
/// Each component (`target`, `span`, `field`, `value`, and `level`) will be covered in turn.
///
/// - `target` matches the event or span's target. In general, this is the module path and/or crate name.
///   Examples of targets `h2`, `tokio::net`, or `tide::server`. For more information on targets,
///   please refer to [`Metadata`]'s documentation.
/// - `span` matches on the span's name. If a `span` directive is provided alongside a `target`,
///   the `span` directive will match on spans _within_ the `target`.
/// - `field` matches on [fields] within spans. Field names can also be supplied without a `value`
///   and will match on any [`Span`] or [`Event`] that has a field with that name.
///   For example: `[span{field=\"value\"}]=debug`, `[{field}]=trace`.
/// - `value` matches on the value of a span's field. If a value is a numeric literal or a bool,
///   it will match _only_ on that value. Otherwise, this filter matches the
///   [`std::fmt::Debug`] output from the value.
/// - `level` sets a maximum verbosity level accepted by this directive.
///
/// When a field value directive (`[{<FIELD NAME>=<FIELD_VALUE>}]=...`) matches a
/// value's [`std::fmt::Debug`] output (i.e., the field value in the directive
/// is not a `bool`, `i64`, `u64`, or `f64` literal), the matched pattern may be
/// interpreted as either a regular expression or as the precise expected
/// output of the field's [`std::fmt::Debug`] implementation. By default, these
/// filters are interpreted as regular expressions, but this can be disabled
/// using the [`Builder::with_regex`] builder method to use precise matching
/// instead.
///
/// When field value filters are interpreted as regular expressions, the
/// [`regex` crate's regular expression syntax][re-syntax] is supported.
///
/// **Note**: When filters are constructed from potentially untrusted inputs,
/// [disabling regular expression matching](Builder::with_regex) is strongly
/// recommended.
///
/// ## Usage Notes
///
/// - The portion of the directive which is included within the square brackets is `tracing`-specific.
/// - Any portion of the directive can be omitted.
///     - The sole exception are the `field` and `value` directives. If a `value` is provided,
///       a `field` must _also_ be provided. However, the converse does not hold, as fields can
///       be matched without a value.
/// - If only a level is provided, it will set the maximum level for all `Span`s and `Event`s
///   that are not enabled by other filters.
/// - A directive without a level will enable anything that it matches. This is equivalent to `=trace`.
/// - When a crate has a dash in its name, the default target for events will be the
///   crate's module path as it appears in Rust. This means every dash will be replaced
///   with an underscore.
/// - A dash in a target will only appear when being specified explicitly:
///   `tracing::info!(target: "target-name", ...);`
///
/// ## Example Syntax
///
/// - `tokio::net=info` will enable all spans or events that:
///    - have the `tokio::net` target,
///    - at the level `info` or above.
/// - `warn,tokio::net=info` will enable all spans and events that:
///    - are at the level `warn` or above, *or*
///    - have the `tokio::net` target at the level `info` or above.
/// - `my_crate[span_a]=trace` will enable all spans and events that:
///    - are within the `span_a` span or named `span_a` _if_ `span_a` has the target `my_crate`,
///    - at the level `trace` or above.
/// - `[span_b{name=\"bob\"}]` will enable all spans or event that:
///    - have _any_ target,
///    - are inside a span named `span_b`,
///    - which has a field named `name` with value `bob`,
///    - at _any_ level.
///
/// # Examples
///
/// Parsing an `EnvFilter` from the [default environment
/// variable](EnvFilter::from_default_env) (`RUST_LOG`):
///
/// ```
/// use tracing_subscriber::{EnvFilter, fmt, prelude::*};
///
/// tracing_subscriber::registry()
///     .with(fmt::layer())
///     .with(EnvFilter::from_default_env())
///     .init();
/// ```
///
/// Parsing an `EnvFilter` [from a user-provided environment
/// variable](EnvFilter::from_env):
///
/// ```
/// use tracing_subscriber::{EnvFilter, fmt, prelude::*};
///
/// tracing_subscriber::registry()
///     .with(fmt::layer())
///     .with(EnvFilter::from_env("MYAPP_LOG"))
///     .init();
/// ```
///
/// Using `EnvFilter` as a [per-layer filter][plf] to filter only a single
/// [`Layer`]:
///
/// ```
/// use tracing_subscriber::{EnvFilter, fmt, prelude::*};
///
/// // Parse an `EnvFilter` configuration from the `RUST_LOG`
/// // environment variable.
/// let filter = EnvFilter::from_default_env();
///
/// // Apply the filter to this layer *only*.
/// let filtered_layer = fmt::layer().with_filter(filter);
///
/// // Some other layer, whose output we don't want to filter.
/// let unfiltered_layer = // ...
///     # fmt::layer();
///
/// tracing_subscriber::registry()
///     .with(filtered_layer)
///     .with(unfiltered_layer)
///     .init();
/// ```
/// # Constructing `EnvFilter`s
///
/// An `EnvFilter` is be constructed by parsing a string containing one or more
/// directives. The [`EnvFilter::new`] constructor parses an `EnvFilter` from a
/// string, ignoring any invalid directives, while [`EnvFilter::try_new`]
/// returns an error if invalid directives are encountered. Similarly, the
/// [`EnvFilter::from_env`] and [`EnvFilter::try_from_env`] constructors parse
/// an `EnvFilter` from the value of the provided environment variable, with
/// lossy and strict validation, respectively.
///
/// A [builder](EnvFilter::builder) interface is available to set additional
/// configuration options prior to parsing an `EnvFilter`. See the [`Builder`
/// type's documentation](Builder) for details on the options that can be
/// configured using the builder.
///
/// [`Span`]: tracing_core::span
/// [fields]: tracing_core::Field
/// [`Event`]: tracing_core::Event
/// [`level`]: tracing_core::Level
/// [`Metadata`]: tracing_core::Metadata
/// [`Targets`]: crate::filter::Targets
/// [`env_logger`]: https://crates.io/crates/env_logger
/// [`Filter`]: #impl-Filter<S>
/// [global]: crate::layer#global-filtering
/// [plf]: crate::layer#per-layer-filtering
/// [filtering]: crate::layer#filtering-with-layers
/// [re-syntax]: https://docs.rs/regex/1.11.1/regex/#syntax
#[cfg_attr(docsrs, doc(cfg(all(feature = "env-filter", feature = "std"))))]
#[derive(Debug)]
pub struct EnvFilter {
    statics: directive::Statics,
    dynamics: directive::Dynamics,
    has_dynamics: bool,
    by_id: RwLock<HashMap<span::Id, directive::SpanMatcher>>,
    by_cs: RwLock<HashMap<callsite::Identifier, directive::CallsiteMatcher>>,
    scope: ThreadLocal<RefCell<Vec<LevelFilter>>>,
    regex: bool,
}

/// Creates an [`EnvFilter`] with the same directives as `self`.
///
/// This does *not* clone any of the dynamic state that [`EnvFilter`] acquires while attached to a
/// subscriber.
impl Clone for EnvFilter {
    fn clone(&self) -> EnvFilter {
        EnvFilter {
            statics: self.statics.clone(),
            dynamics: self.dynamics.clone(),
            has_dynamics: self.has_dynamics,
            by_id: RwLock::default(),
            by_cs: RwLock::default(),
            scope: ThreadLocal::new(),
            regex: self.regex,
        }
    }
}

type FieldMap<T> = HashMap<Field, T>;

/// Indicates that an error occurred while parsing a `EnvFilter` from an
/// environment variable.
#[cfg_attr(docsrs, doc(cfg(all(feature = "env-filter", feature = "std"))))]
#[derive(Debug)]
pub struct FromEnvError {
    kind: ErrorKind,
}

#[derive(Debug)]
enum ErrorKind {
    Parse(ParseError),
    Env(env::VarError),
}

impl EnvFilter {
    /// `RUST_LOG` is the default environment variable used by
    /// [`EnvFilter::from_default_env`] and [`EnvFilter::try_from_default_env`].
    ///
    /// [`EnvFilter::from_default_env`]: EnvFilter::from_default_env()
    /// [`EnvFilter::try_from_default_env`]: EnvFilter::try_from_default_env()
    pub const DEFAULT_ENV: &'static str = "RUST_LOG";

    // === constructors, etc ===

    /// Returns a [builder] that can be used to configure a new [`EnvFilter`]
    /// instance.
    ///
    /// The [`Builder`] type is used to set additional configurations, such as
    /// [whether regular expressions are enabled](Builder::with_regex) or [the
    /// default directive](Builder::with_default_directive) before parsing an
    /// [`EnvFilter`] from a string or environment variable.
    ///
    /// [builder]: https://rust-unofficial.github.io/patterns/patterns/creational/builder.html
    pub fn builder() -> Builder {
        Builder::default()
    }

    /// Returns a new `EnvFilter` from the value of the `RUST_LOG` environment
    /// variable, ignoring any invalid filter directives.
    ///
    /// If the environment variable is empty or not set, or if it contains only
    /// invalid directives, a default directive enabling the [`ERROR`] level is
    /// added.
    ///
    /// To set additional configuration options prior to parsing the filter, use
    /// the [`Builder`] type instead.
    ///
    /// This function is equivalent to the following:
    ///
    /// ```rust
    /// use tracing_subscriber::filter::{EnvFilter, LevelFilter};
    ///
    /// # fn docs() -> EnvFilter {
    /// EnvFilter::builder()
    ///     .with_default_directive(LevelFilter::ERROR.into())
    ///     .from_env_lossy()
    /// # }
    /// ```
    ///
    /// [`ERROR`]: tracing::Level::ERROR
    pub fn from_default_env() -> Self {
        Self::builder()
            .with_default_directive(LevelFilter::ERROR.into())
            .from_env_lossy()
    }

    /// Returns a new `EnvFilter` from the value of the given environment
    /// variable, ignoring any invalid filter directives.
    ///
    /// If the environment variable is empty or not set, or if it contains only
    /// invalid directives, a default directive enabling the [`ERROR`] level is
    /// added.
    ///
    /// To set additional configuration options prior to parsing the filter, use
    /// the [`Builder`] type instead.
    ///
    /// This function is equivalent to the following:
    ///
    /// ```rust
    /// use tracing_subscriber::filter::{EnvFilter, LevelFilter};
    ///
    /// # fn docs() -> EnvFilter {
    /// # let env = "";
    /// EnvFilter::builder()
    ///     .with_default_directive(LevelFilter::ERROR.into())
    ///     .with_env_var(env)
    ///     .from_env_lossy()
    /// # }
    /// ```
    ///
    /// [`ERROR`]: tracing::Level::ERROR
    pub fn from_env<A: AsRef<str>>(env: A) -> Self {
        Self::builder()
            .with_default_directive(LevelFilter::ERROR.into())
            .with_env_var(env.as_ref())
            .from_env_lossy()
    }

    /// Returns a new `EnvFilter` from the directives in the given string,
    /// ignoring any that are invalid.
    ///
    /// If the string is empty or contains only invalid directives, a default
    /// directive enabling the [`ERROR`] level is added.
    ///
    /// To set additional configuration options prior to parsing the filter, use
    /// the [`Builder`] type instead.
    ///
    /// This function is equivalent to the following:
    ///
    /// ```rust
    /// use tracing_subscriber::filter::{EnvFilter, LevelFilter};
    ///