Files
logger/README.md
T
mifi f464fa5017
ci/woodpecker/push/ci Pipeline failed
ci/woodpecker/push/publish unknown status
fix(docs):Docs and Exports
- Add documentation for recently released changes
- Update some exports to include enums
2026-08-21 11:45:56 -03:00

246 lines
11 KiB
Markdown
Raw Blame History

This file contains ambiguous Unicode characters
This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.
# @mifi/logger
Intentional, namespaced logging for TypeScript browser and Node.js applications.
```ts
import { createLogger } from '@mifi/logger';
const logger = createLogger({ environment: 'production', namespace: 'WIDGET' });
logger.child('Button').debug('Rendered', () => ({ expensive: 'only evaluated when shown' }));
logger.debug('response', async () => ({ data: await fetchBody() })); // settled before sinks
```
## Install
```bash
pnpm add @mifi/logger
```
Requires Node.js `>=24`. Published to the private `@mifi` registry.
## Quick start
```ts
import { createLogger } from '@mifi/logger';
const logger = createLogger({
environment: 'development', // "development" | "staging" | "production"
namespace: 'API', // optional; string or string[] joined with ":"
});
logger.trace('very detailed');
logger.debug('diagnostic', () => ({ snapshot: heavyWork() }));
logger.debug('response', async () => ({ data: await res.json(), url: res.url }));
logger.info('ready');
logger.warn('slow response', { ms: 1200 });
logger.error('request failed', error);
logger.assert(userId, 'missing user id');
const auth = logger.child('auth'); // namespace → "API:auth"
auth.info('token refreshed');
```
Console utility methods (`group`, `groupCollapsed`, `table`, `time`, `count`, `dir`, …) are available and gated by the same level policy as `info`.
## Runtime policy
Default levels by environment:
| Environment | Default level | Console shows |
| ------------- | ------------- | --------------- |
| `development` | `trace` | all levels |
| `staging` | `warn` | `warn`, `error` |
| `production` | `error` | `error` only |
Levels in order (most → least verbose): `trace``debug``info``warn``error``silent`.
### Override precedence
Highest wins:
1. Explicit `level` option on `createLogger`
2. Browser `sessionStorage.showLoggingFor`
3. Env vars `MIFI_LOG_LEVEL` / `MIFI_LOG_NAMESPACES`
4. Environment default (table above)
### Browser session override
```js
sessionStorage.setItem('showLoggingFor', 'debug');
// or restrict to namespaces (exact or descendants):
sessionStorage.setItem('showLoggingFor', 'debug:WIDGET,API');
```
`WIDGET` matches `WIDGET` and `WIDGET:Button`.
### Container / Node env overrides
See [Environment variables](#environment-variables) and [`env.example`](./env.example).
```bash
MIFI_LOG_LEVEL=debug
MIFI_LOG_NAMESPACES=WIDGET,API
```
## Environment variables
These are read from `process.env` (or from `options.env` when testing/injecting):
| Variable | Description |
| --------------------- | --------------------------------------------------------------------------- |
| `MIFI_LOG_LEVEL` | Override level: `trace`, `debug`, `info`, `warn`, `error`, or `silent`. |
| `MIFI_LOG_NAMESPACES` | Optional comma-separated namespace filters applied with the level override. |
When only `MIFI_LOG_LEVEL` is set, all namespaces are enabled at that level. When `MIFI_LOG_NAMESPACES` is also set, only matching namespaces (and their children) emit at that level.
Copy [`env.example`](./env.example) as a starting point for local or container config.
## Sentry
The Sentry adapter has **no SDK dependency**. Pass your official `@sentry/node`, `@sentry/browser`, or `@sentry/nextjs` module directly to `createSentrySink` — no consumer adapter or cast is required.
**Development never sends to Sentry.** Staging and production attach the sink when `sentry` is configured.
Destination policy when `sentry` is configured and `sinks` is not:
| Runtime + environment | Console | Sentry |
| ---------------------- | ---------------------- | -------------------------------- |
| Production **browser** | Off | Issues for errors; optional Logs |
| Production **Node** | On (stderr for errors) | Issues for errors; optional Logs |
| Staging | On (default `warn`+) | Issues for errors; optional Logs |
| Development | On | Nothing |
### Issues vs Logs
- **Errors** → Sentry Issues (`captureException` / `captureMessage`). Not also written to Logs.
- **Warnings and lower** → Sentry Logs (`sentry.logger.*`) only when `{ logs: true }` on the sink (or forced per call).
Default Sentry Logs thresholds when `logs` is enabled:
| Environment | Logs threshold |
| ----------- | -------------- |
| Staging | `info` |
| Production | `warn` |
Override with `logLevel` on `createSentrySink`. Console policy stays independent — production browsers can stay quiet while warnings still appear in Sentry Logs.
### Per-call options
`trace` / `debug` / `log` / `info` / `warn` / `error` (and `assert`) accept a trailing options object. Console utilities (`group`, `time`, …) do not.
```ts
import * as Sentry from '@sentry/nextjs';
import { createLogger } from '@mifi/logger';
import { createSentrySink } from '@mifi/logger/sentry';
const logger = createLogger({
environment: 'production',
namespace: 'API',
sentry: createSentrySink(Sentry, { logs: true }),
});
logger.error('Request failed', error, { requestId: 'req_1' });
logger.warn('slow', { ms: 1200 }); // → Sentry Logs in production
logger.info('investigating', { orderId }, { sentry: true }); // force Logs
logger.error('expected', err, { suppressSentry: true }); // console only (when enabled)
```
| Option | Effect |
| ---------------------- | ------------------------------------------------------------------------ |
| `sentry: true` | Force this event to Sentry Logs (ignores Logs threshold; not for errors) |
| `suppressSentry: true` | Skip creating a Sentry Issue for an error |
A last argument is treated as options only when every key is `sentry` or `suppressSentry`. Prefer the third-argument form when you also pass data.
Logs payload: string message + flat primitive attributes (including `logger.namespace`). Nested objects are not unfurled. When emit was deferred for thenables, the payload also includes `async: true`.
Issue severities use Sentrys union (`warning`, not logger `warn`). The sink maps `warn``warning` and `trace``debug` for Issues only; Logs keep logger level names (`sentry.logger.warn`).
## Custom sinks
Replace the default destinations entirely with `sinks`:
```ts
import { createLogger, createConsoleSink, type LogSink } from '@mifi/logger';
const analyticsSink: LogSink = {
emit(event) {
if (event.sendToConsole) analytics.track('log', event);
// event.async === true when args were settled asynchronously
},
};
const logger = createLogger({
environment: 'production',
sinks: [createConsoleSink(), analyticsSink],
});
```
When `sinks` is provided, the `sentry` option is not used for default sink selection.
## API overview
| Export | From | Purpose |
| ---------------------- | --------------------- | ------------------------------------------------ |
| `createLogger` | `@mifi/logger` | Build a namespaced, Console-shaped logger |
| `createConsoleSink` | `@mifi/logger` | Default styled console destination |
| `parseLoggingOverride` | `@mifi/logger` | Parse `debug` / `debug:NS1,NS2` override strings |
| `createSentrySink` | `@mifi/logger/sentry` | Adapter for any Sentry-like SDK |
Types: `Logger`, `LoggerOptions`, `LogLevel`, `LoggerEnvironment`, `LogEvent`, `LogSink`, `LogData`, `LogCallOptions`, and (from `/sentry`) `SentryLike`, `SentryScopeLike`, `SentrySeverityLevel`, `SentrySink`, `SentrySinkOptions`.
Optional destinations live under `src/sinks/` (console, Sentry today; New Relic / Datadog-style adapters can land beside them later). Public import paths stay `@mifi/logger` and `@mifi/logger/sentry`.
All public APIs include JSDoc with parameter and example documentation in the TypeScript sources.
## Lazy arguments
Any function passed **after** the first argument is invoked only if the event will actually be emitted (console and/or Sentry). That keeps expensive snapshots cheap when the level is suppressed.
```ts
logger.debug('state', () => buildHugeSnapshot()); // skipped when debug is suppressed
logger.error('failed', error, () => ({ body: pendingBody }));
```
### Async factories and Promises
Factories may be `async`, or you may pass a top-level `Promise` as an argument. Thenables are **settled before sinks run**; the log method stays fire-and-forget (`void`).
```ts
logger.debug('response', async () => ({
data: await res.json(),
url: res.url,
}));
logger.error('failed', { id }, responsePromise);
```
| Behavior | Detail |
| ----------------- | -------------------------------------------------------------------------------------------- |
| Unwrap scope | **Shallow** — only top-level thenables in the argument list |
| Nested promises | Not walked; compose with `await` / `Promise.all` inside the factory |
| Call return type | Still `void` (no `await logger.debug(...)`) |
| Timestamp | Captured at call time (before settle), so async logs correlate with the failure moment |
| `LogEvent.async` | `true` when emit was deferred; omitted for fully sync emits |
| Console | Label gains an `[async]` suffix, e.g. `[API][async]` |
| Sentry Logs | Attribute `async: true` when set |
| Ordering | Async emits may appear after later sync logs |
| Rejections | Rejection `reason` replaces that argument slot; other args still emit (`Promise.allSettled`) |
| Console utilities | `group` / `time` / etc. do not settle thenables (emit path only) |
```ts
// Nested promises — compose inside the factory:
logger.debug('route', async () => {
const [body, next] = await Promise.all([res.json(), nextRoute]);
return { body, next };
});
```
## Releases
Woodpecker verifies pull requests and `main`, reports CI to Mattermost, then automatically releases from Conventional Commits merged to `main`. It updates `package.json` and `CHANGELOG.md`, creates a `vX.Y.Z` tag, and publishes to the private `@mifi` registry.
Use `fix:` for a patch, `feat:` for a minor release, and `feat!` / `fix!` (or a `BREAKING CHANGE:` footer) for a major release. Commits such as `docs:`, `test:`, and `chore:` do not release a package.
The release workflow uses the existing global `gitea_package_token`, `gitea_registry_username`, `gitea_release_token`, `mattermost_bot_access_token`, `mattermost_tests_channel_id`, `mattermost_pushes_channel_id`, and `mattermost_post_api_url` secrets.