temporalio 0.1.1 → 0.3.0

data/bridge/sdk-core/core-api/Cargo.toml

@@ -1,33 +0,0 @@
-[package]
-name = "temporal-sdk-core-api"
-version = "0.1.0"
-edition = "2021"
-authors = ["Spencer Judge <spencer@temporal.io>"]
-license-file = "LICENSE.txt"
-description = "Interface definitions for the Temporal Core SDK"
-homepage = "https://temporal.io/"
-repository = "https://github.com/temporalio/sdk-core"
-keywords = ["temporal", "workflow"]
-categories = ["development-tools"]
-
-# See more keys and their definitions at https://doc.rust-lang.org/cargo/reference/manifest.html
-
-[dependencies]
-async-trait = "0.1"
-derive_builder = "0.12"
-prost-types = "0.11"
-serde = { version = "1.0", default_features = false, features = ["derive"] }
-serde_json = "1.0"
-thiserror = "1.0"
-tokio = "1.24"
-tonic = "0.8"
-tracing-core = "0.1"
-url = "2.3"
-
-[dependencies.temporal-sdk-core-protos]
-path = "../sdk-core-protos"
-version = "0.1"
-
-[dependencies.temporal-client]
-path = "../client"
-version = "0.1"

data/bridge/sdk-core/core-api/LICENSE.txt

@@ -1,23 +0,0 @@
-Temporal Core SDK
-
-The MIT License
-
-Copyright (c) 2021 Temporal Technologies, Inc. All Rights Reserved
-
-Permission is hereby granted, free of charge, to any person obtaining a copy
-of this software and associated documentation files (the "Software"), to deal
-in the Software without restriction, including without limitation the rights
-to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
-copies of the Software, and to permit persons to whom the Software is
-furnished to do so, subject to the following conditions:
-
-The above copyright notice and this permission notice shall be included in all
-copies or substantial portions of the Software.
-
-THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
-IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
-FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
-AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
-LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
-OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
-SOFTWARE.

data/bridge/sdk-core/core-api/src/errors.rs

@@ -1,62 +0,0 @@
-//! Error types exposed by public APIs
-
-use temporal_sdk_core_protos::coresdk::activity_result::ActivityExecutionResult;
-
-/// Errors thrown by [crate::Worker::poll_workflow_activation]
-#[derive(thiserror::Error, Debug)]
-pub enum PollWfError {
-    /// [crate::Worker::shutdown] was called, and there are no more replay tasks to be handled. Lang
-    /// must call [crate::Worker::complete_workflow_activation] for any remaining tasks, and then may
-    /// exit.
-    #[error("Core is shut down and there are no more workflow replay tasks")]
-    ShutDown,
-    /// Unhandled error when calling the temporal server. Core will attempt to retry any non-fatal
-    /// errors, so lang should consider this fatal.
-    #[error("Unhandled grpc error when workflow polling: {0:?}")]
-    TonicError(#[from] tonic::Status),
-    /// Unhandled error when completing a workflow during a poll -- this can happen when there is no
-    /// work for lang to perform, but the server sent us a workflow task (EX: An activity completed
-    /// even though we already cancelled it)
-    #[error("Unhandled error when auto-completing workflow task: {0:?}")]
-    AutocompleteError(#[from] CompleteWfError),
-}
-
-/// Errors thrown by [crate::Worker::poll_activity_task]
-#[derive(thiserror::Error, Debug)]
-pub enum PollActivityError {
-    /// [crate::Worker::shutdown] was called, we will no longer fetch new activity tasks. Lang must
-    /// ensure it is finished with any workflow replay, see [PollWfError::ShutDown]
-    #[error("Core is shut down")]
-    ShutDown,
-    /// Unhandled error when calling the temporal server. Core will attempt to retry any non-fatal
-    /// errors, so lang should consider this fatal.
-    #[error("Unhandled grpc error when activity polling: {0:?}")]
-    TonicError(#[from] tonic::Status),
-}
-
-/// Errors thrown by [crate::Worker::complete_workflow_activation]
-#[derive(thiserror::Error, Debug)]
-#[allow(clippy::large_enum_variant)]
-pub enum CompleteWfError {
-    /// Lang SDK sent us a malformed workflow completion. This likely means a bug in the lang sdk.
-    #[error("Lang SDK sent us a malformed workflow completion for run ({run_id}): {reason}")]
-    MalformedWorkflowCompletion {
-        /// Reason the completion was malformed
-        reason: String,
-        /// The run associated with the completion
-        run_id: String,
-    },
-}
-
-/// Errors thrown by [crate::Worker::complete_activity_task]
-#[derive(thiserror::Error, Debug)]
-pub enum CompleteActivityError {
-    /// Lang SDK sent us a malformed activity completion. This likely means a bug in the lang sdk.
-    #[error("Lang SDK sent us a malformed activity completion ({reason}): {completion:?}")]
-    MalformedActivityCompletion {
-        /// Reason the completion was malformed
-        reason: String,
-        /// The completion, which may not be included to avoid unnecessary copies.
-        completion: Option<ActivityExecutionResult>,
-    },
-}

data/bridge/sdk-core/core-api/src/lib.rs

@@ -1,113 +0,0 @@
-pub mod errors;
-pub mod telemetry;
-pub mod worker;
-
-use crate::{
-    errors::{CompleteActivityError, CompleteWfError, PollActivityError, PollWfError},
-    worker::WorkerConfig,
-};
-use temporal_sdk_core_protos::coresdk::{
-    activity_task::ActivityTask, workflow_activation::WorkflowActivation,
-    workflow_completion::WorkflowActivationCompletion, ActivityHeartbeat, ActivityTaskCompletion,
-};
-
-/// This trait is the primary way by which language specific SDKs interact with the core SDK.
-/// It represents one worker, which has a (potentially shared) client for connecting to the service
-/// and is bound to a specific task queue.
-#[async_trait::async_trait]
-pub trait Worker: Send + Sync {
-    /// Ask the worker for some work, returning a [WorkflowActivation]. It is then the language
-    /// SDK's responsibility to call the appropriate workflow code with the provided inputs. Blocks
-    /// indefinitely until such work is available or [Worker::shutdown] is called.
-    ///
-    /// It is important to understand that all activations must be responded to. There can only
-    /// be one outstanding activation for a particular run of a workflow at any time. If an
-    /// activation is not responded to, it will cause that workflow to become stuck forever.
-    ///
-    /// Activations that contain only a `remove_from_cache` job should not cause the workflow code
-    /// to be invoked and may be responded to with an empty command list. Eviction jobs may also
-    /// appear with other jobs, but will always appear last in the job list. In this case it is
-    /// expected that the workflow code will be invoked, and the response produced as normal, but
-    /// the caller should evict the run after doing so.
-    ///
-    /// It is rarely a good idea to call poll concurrently. It handles polling the server
-    /// concurrently internally.
-    async fn poll_workflow_activation(&self) -> Result<WorkflowActivation, PollWfError>;
-
-    /// Ask the worker for some work, returning an [ActivityTask]. It is then the language SDK's
-    /// responsibility to call the appropriate activity code with the provided inputs. Blocks
-    /// indefinitely until such work is available or [Worker::shutdown] is called.
-    ///
-    /// The returned activation is guaranteed to be for the same task queue / worker which was
-    /// provided as the `task_queue` argument.
-    ///
-    /// It is rarely a good idea to call poll concurrently. It handles polling the server
-    /// concurrently internally.
-    async fn poll_activity_task(&self) -> Result<ActivityTask, PollActivityError>;
-
-    /// Tell the worker that a workflow activation has completed. May (and should) be freely called
-    /// concurrently. The future may take some time to resolve, as fetching more events might be
-    /// necessary for completion to... complete - thus SDK implementers should make sure they do
-    /// not serialize completions.
-    async fn complete_workflow_activation(
-        &self,
-        completion: WorkflowActivationCompletion,
-    ) -> Result<(), CompleteWfError>;
-
-    /// Tell the worker that an activity has finished executing. May (and should) be freely called
-    /// concurrently.
-    async fn complete_activity_task(
-        &self,
-        completion: ActivityTaskCompletion,
-    ) -> Result<(), CompleteActivityError>;
-
-    /// Notify the Temporal service that an activity is still alive. Long running activities that
-    /// take longer than `activity_heartbeat_timeout` to finish must call this function in order to
-    /// report progress, otherwise the activity will timeout and a new attempt will be scheduled.
-    ///
-    /// The first heartbeat request will be sent immediately, subsequent rapid calls to this
-    /// function will result in heartbeat requests being aggregated and the last one received during
-    /// the aggregation period will be sent to the server, where that period is defined as half the
-    /// heartbeat timeout.
-    ///
-    /// Unlike Java/Go SDKs we do not return cancellation status as part of heartbeat response and
-    /// instead send it as a separate activity task to the lang, decoupling heartbeat and
-    /// cancellation processing.
-    ///
-    /// For now activity still need to send heartbeats if they want to receive cancellation
-    /// requests. In the future we will change this and will dispatch cancellations more
-    /// proactively. Note that this function does not block on the server call and returns
-    /// immediately. Underlying validation errors are swallowed and logged, this has been agreed to
-    /// be optimal behavior for the user as we don't want to break activity execution due to badly
-    /// configured heartbeat options.
-    fn record_activity_heartbeat(&self, details: ActivityHeartbeat);
-
-    /// Request that a workflow be evicted by its run id. This will generate a workflow activation
-    /// with the eviction job inside it to be eventually returned by
-    /// [Worker::poll_workflow_activation]. If the workflow had any existing outstanding activations,
-    /// such activations are invalidated and subsequent completions of them will do nothing and log
-    /// a warning.
-    fn request_workflow_eviction(&self, run_id: &str);
-
-    /// Return this worker's config
-    fn get_config(&self) -> &WorkerConfig;
-
-    /// Initiate shutdown. See [Worker::shutdown], this is just a sync version that starts the
-    /// process. You can then wait on `shutdown` or [Worker::finalize_shutdown].
-    fn initiate_shutdown(&self);
-
-    /// Initiates async shutdown procedure, eventually ceases all polling of the server and shuts
-    /// down this worker. [Worker::poll_workflow_activation] should be called until it
-    /// returns [PollWfError::ShutDown] to ensure that any workflows which are still undergoing
-    /// replay have an opportunity to finish. This means that the lang sdk will need to call
-    /// [Worker::complete_workflow_activation] for those workflows until they are done. At that point,
-    /// the lang SDK can end the process, or drop the [Worker] instance, which will close the
-    /// connection.
-    async fn shutdown(&self);
-
-    /// Completes shutdown and frees all resources. You should avoid simply dropping workers, as
-    /// this does not allow async tasks to report any panics that may have occurred cleanly.
-    ///
-    /// This should be called only after [Worker::shutdown] has resolved.
-    async fn finalize_shutdown(self);
-}

data/bridge/sdk-core/core-api/src/telemetry.rs

@@ -1,141 +0,0 @@
-use std::{
-    collections::HashMap,
-    net::SocketAddr,
-    time::{Duration, SystemTime, UNIX_EPOCH},
-};
-use tracing_core::Level;
-use url::Url;
-
-/// Each core runtime instance has a telemetry subsystem associated with it, this trait defines the
-/// operations that lang might want to perform on that telemetry after it's initialized.
-pub trait CoreTelemetry {
-    /// Each worker buffers logs that should be shuttled over to lang so that they may be rendered
-    /// with the user's desired logging library. Use this function to grab the most recent buffered
-    /// logs since the last time it was called. A fixed number of such logs are retained at maximum,
-    /// with the oldest being dropped when full.
-    ///
-    /// Returns the list of logs from oldest to newest. Returns an empty vec if the feature is not
-    /// configured.
-    fn fetch_buffered_logs(&self) -> Vec<CoreLog>;
-}
-
-/// Telemetry configuration options. Construct with [TelemetryOptionsBuilder]
-#[derive(Debug, Clone, derive_builder::Builder)]
-#[non_exhaustive]
-pub struct TelemetryOptions {
-    /// Optional trace exporter - set as None to disable.
-    #[builder(setter(into, strip_option), default)]
-    pub tracing: Option<TraceExportConfig>,
-    /// Optional logger - set as None to disable.
-    #[builder(setter(into, strip_option), default)]
-    pub logging: Option<Logger>,
-    /// Optional metrics exporter - set as None to disable.
-    #[builder(setter(into, strip_option), default)]
-    pub metrics: Option<MetricsExporter>,
-
-    /// If set true, do not prefix metrics with `temporal_`. Will be removed eventually as
-    /// the prefix is consistent with other SDKs.
-    #[builder(default)]
-    pub no_temporal_prefix_for_metrics: bool,
-
-    /// Specifies the aggregation temporality for metric export. Defaults to cumulative.
-    #[builder(default = "MetricTemporality::Cumulative")]
-    pub metric_temporality: MetricTemporality,
-}
-
-/// Options for exporting to an OpenTelemetry Collector
-#[derive(Debug, Clone)]
-pub struct OtelCollectorOptions {
-    /// The url of the OTel collector to export telemetry and metrics to. Lang SDK should also
-    /// export to this same collector.
-    pub url: Url,
-    /// Optional set of HTTP headers to send to the Collector, e.g for authentication.
-    pub headers: HashMap<String, String>,
-    /// Optionally specify how frequently metrics should be exported. Defaults to 1 second.
-    pub metric_periodicity: Option<Duration>,
-}
-
-/// Configuration for the external export of traces
-#[derive(Debug, Clone)]
-pub struct TraceExportConfig {
-    /// An [EnvFilter](https://docs.rs/tracing-subscriber/latest/tracing_subscriber/struct.EnvFilter.html) filter string.
-    pub filter: String,
-    /// Where they should go
-    pub exporter: TraceExporter,
-}
-
-/// Control where traces are exported.
-#[derive(Debug, Clone)]
-pub enum TraceExporter {
-    /// Export traces to an OpenTelemetry Collector <https://opentelemetry.io/docs/collector/>.
-    Otel(OtelCollectorOptions),
-}
-
-/// Control where metrics are exported
-#[derive(Debug, Clone)]
-pub enum MetricsExporter {
-    /// Export metrics to an OpenTelemetry Collector <https://opentelemetry.io/docs/collector/>.
-    Otel(OtelCollectorOptions),
-    /// Expose metrics directly via an embedded http server bound to the provided address.
-    Prometheus(SocketAddr),
-}
-
-/// Control where logs go
-#[derive(Debug, Clone)]
-pub enum Logger {
-    /// Log directly to console.
-    Console {
-        /// An [EnvFilter](https://docs.rs/tracing-subscriber/latest/tracing_subscriber/struct.EnvFilter.html) filter string.
-        filter: String,
-    },
-    /// Forward logs to Lang - collectable with `fetch_global_buffered_logs`.
-    Forward {
-        /// An [EnvFilter](https://docs.rs/tracing-subscriber/latest/tracing_subscriber/struct.EnvFilter.html) filter string.
-        filter: String,
-    },
-}
-
-/// Types of aggregation temporality for metric export.
-/// See: <https://github.com/open-telemetry/opentelemetry-specification/blob/ce50e4634efcba8da445cc23523243cb893905cb/specification/metrics/datamodel.md#temporality>
-#[derive(Debug, Clone, Copy)]
-pub enum MetricTemporality {
-    /// Successive data points repeat the starting timestamp
-    Cumulative,
-    /// Successive data points advance the starting timestamp
-    Delta,
-}
-
-impl Default for TelemetryOptions {
-    fn default() -> Self {
-        TelemetryOptionsBuilder::default().build().unwrap()
-    }
-}
-
-/// A log line (which ultimately came from a tracing event) exported from Core->Lang
-#[derive(Debug)]
-pub struct CoreLog {
-    /// The module within core this message originated from
-    pub target: String,
-    /// Log message
-    pub message: String,
-    /// Time log was generated (not when it was exported to lang)
-    pub timestamp: SystemTime,
-    /// Message level
-    pub level: Level,
-    /// Arbitrary k/v pairs (span k/vs are collapsed with event k/vs here). We could change this
-    /// to include them in `span_contexts` instead, but there's probably not much value for log
-    /// forwarding.
-    pub fields: HashMap<String, serde_json::Value>,
-    /// A list of the outermost to the innermost span names
-    pub span_contexts: Vec<String>,
-}
-
-impl CoreLog {
-    /// Return timestamp as ms since epoch
-    pub fn millis_since_epoch(&self) -> u128 {
-        self.timestamp
-            .duration_since(UNIX_EPOCH)
-            .unwrap_or(Duration::ZERO)
-            .as_millis()
-    }
-}

data/bridge/sdk-core/core-api/src/worker.rs

@@ -1,161 +0,0 @@
-use std::time::Duration;
-use tokio::sync::mpsc::UnboundedSender;
-
-/// Defines per-worker configuration options
-#[derive(Debug, Clone, derive_builder::Builder, serde::Serialize, serde::Deserialize)]
-#[builder(setter(into), build_fn(validate = "Self::validate"))]
-#[non_exhaustive]
-pub struct WorkerConfig {
-    /// The Temporal service namespace this worker is bound to
-    pub namespace: String,
-    /// What task queue will this worker poll from? This task queue name will be used for both
-    /// workflow and activity polling.
-    pub task_queue: String,
-    /// A string that should be unique to the set of code this worker uses. IE: All the workflow,
-    /// activity, interceptor, and data converter code.
-    pub worker_build_id: String,
-    /// A human-readable string that can identify this worker. Using something like sdk version
-    /// and host name is a good default. If set, overrides the identity set (if any) on the client
-    /// used by this worker.
-    #[builder(default)]
-    pub client_identity_override: Option<String>,
-    /// If set nonzero, workflows will be cached and sticky task queues will be used, meaning that
-    /// history updates are applied incrementally to suspended instances of workflow execution.
-    /// Workflows are evicted according to a least-recently-used policy one the cache maximum is
-    /// reached. Workflows may also be explicitly evicted at any time, or as a result of errors
-    /// or failures.
-    #[builder(default = "0")]
-    pub max_cached_workflows: usize,
-    /// The maximum allowed number of workflow tasks that will ever be given to this worker at one
-    /// time. Note that one workflow task may require multiple activations - so the WFT counts as
-    /// "outstanding" until all activations it requires have been completed.
-    ///
-    /// Cannot be larger than `max_cached_workflows`.
-    #[builder(default = "100")]
-    pub max_outstanding_workflow_tasks: usize,
-    /// The maximum number of activity tasks that will ever be given to this worker concurrently
-    #[builder(default = "100")]
-    pub max_outstanding_activities: usize,
-    /// The maximum number of local activity tasks that will ever be given to this worker
-    /// concurrently
-    #[builder(default = "100")]
-    pub max_outstanding_local_activities: usize,
-    /// Maximum number of concurrent poll workflow task requests we will perform at a time on this
-    /// worker's task queue. See also [WorkerConfig::nonsticky_to_sticky_poll_ratio]. Must be at
-    /// least 1.
-    #[builder(default = "5")]
-    pub max_concurrent_wft_polls: usize,
-    /// [WorkerConfig::max_concurrent_wft_polls] * this number = the number of max pollers that will
-    /// be allowed for the nonsticky queue when sticky tasks are enabled. If both defaults are used,
-    /// the sticky queue will allow 4 max pollers while the nonsticky queue will allow one. The
-    /// minimum for either poller is 1, so if `max_concurrent_wft_polls` is 1 and sticky queues are
-    /// enabled, there will be 2 concurrent polls.
-    #[builder(default = "0.2")]
-    pub nonsticky_to_sticky_poll_ratio: f32,
-    /// Maximum number of concurrent poll activity task requests we will perform at a time on this
-    /// worker's task queue
-    #[builder(default = "5")]
-    pub max_concurrent_at_polls: usize,
-    /// If set to true this worker will only handle workflow tasks and local activities, it will not
-    /// poll for activity tasks.
-    #[builder(default = "false")]
-    pub no_remote_activities: bool,
-    /// How long a workflow task is allowed to sit on the sticky queue before it is timed out
-    /// and moved to the non-sticky queue where it may be picked up by any worker.
-    #[builder(default = "Duration::from_secs(10)")]
-    pub sticky_queue_schedule_to_start_timeout: Duration,
-
-    /// Longest interval for throttling activity heartbeats
-    #[builder(default = "Duration::from_secs(60)")]
-    pub max_heartbeat_throttle_interval: Duration,
-
-    /// Default interval for throttling activity heartbeats in case
-    /// `ActivityOptions.heartbeat_timeout` is unset.
-    /// When the timeout *is* set in the `ActivityOptions`, throttling is set to
-    /// `heartbeat_timeout * 0.8`.
-    #[builder(default = "Duration::from_secs(30)")]
-    pub default_heartbeat_throttle_interval: Duration,
-
-    /// Sets the maximum number of activities per second the task queue will dispatch, controlled
-    /// server-side. Note that this only takes effect upon an activity poll request. If multiple
-    /// workers on the same queue have different values set, they will thrash with the last poller
-    /// winning.
-    #[builder(default)]
-    pub max_task_queue_activities_per_second: Option<f64>,
-
-    /// Limits the number of activities per second that this worker will process. The worker will
-    /// not poll for new activities if by doing so it might receive and execute an activity which
-    /// would cause it to exceed this limit. Negative, zero, or NaN values will cause building
-    /// the options to fail.
-    #[builder(default)]
-    pub max_worker_activities_per_second: Option<f64>,
-
-    /// # UNDER DEVELOPMENT
-    /// If set to true this worker will opt-in to the whole-worker versioning feature.
-    /// `worker_build_id` will be used as the version.
-    /// todo: link to feature docs
-    #[builder(default = "false")]
-    pub use_worker_versioning: bool,
-
-    /// If set false (default), shutdown will not finish until all pending evictions have been
-    /// issued and replied to. If set true shutdown will be considered complete when the only
-    /// remaining work is pending evictions.
-    ///
-    /// This flag is useful during tests to avoid needing to deal with lots of uninteresting
-    /// evictions during shutdown. Alternatively, if a lang implementation finds it easy to clean
-    /// up during shutdown, setting this true saves some back-and-forth.
-    #[builder(default = "false")]
-    pub ignore_evicts_on_shutdown: bool,
-
-    /// Maximum number of next page (or initial) history event listing requests we'll make
-    /// concurrently. I don't this it's worth exposing this to users until we encounter a reason.
-    #[builder(default = "5")]
-    pub fetching_concurrency: usize,
-
-    /// If set, and the `save_wf_inputs` feature is enabled in core, will be sent a serialized
-    /// instance of every input to workflow state in order. This is for testing purposes, SDK
-    /// implementations never need to care about it.
-    #[builder(default)]
-    #[serde(skip)]
-    pub wf_state_inputs: Option<UnboundedSender<Vec<u8>>>,
-}
-
-impl WorkerConfig {
-    pub fn max_nonsticky_polls(&self) -> usize {
-        ((self.max_concurrent_wft_polls as f32 * self.nonsticky_to_sticky_poll_ratio) as usize)
-            .max(1)
-    }
-    pub fn max_sticky_polls(&self) -> usize {
-        self.max_concurrent_wft_polls
-            .saturating_sub(self.max_nonsticky_polls())
-            .max(1)
-    }
-}
-
-impl WorkerConfigBuilder {
-    fn validate(&self) -> Result<(), String> {
-        if self.max_concurrent_wft_polls == Some(0) {
-            return Err("`max_concurrent_wft_polls` must be at least 1".to_owned());
-        }
-        if self.max_concurrent_at_polls == Some(0) {
-            return Err("`max_concurrent_at_polls` must be at least 1".to_owned());
-        }
-        if self.max_cached_workflows > Some(0)
-            && self.max_outstanding_workflow_tasks > self.max_cached_workflows
-        {
-            return Err(
-                "Maximum concurrent workflow tasks cannot exceed the maximum number of cached \
-                 workflows"
-                    .to_owned(),
-            );
-        }
-        if let Some(Some(ref x)) = self.max_worker_activities_per_second {
-            if !x.is_normal() || x.is_sign_negative() {
-                return Err(
-                    "`max_worker_activities_per_second` must be positive and nonzero".to_owned(),
-                );
-            }
-        }
-        Ok(())
-    }
-}