trackplan

Type-safe analytics for Dart and Flutter, generated from one YAML tracking plan. A misspelled event or a missing parameter becomes a compile error.

The same plan also generates TypeScript code with the trackplan npm package, and both tools report the same diagnostics. See the main README and the plan format.

The problem

analytics.logEvent(name: 'chekout_completed', parameters: {'total': '50'});
  • The typo creates a second event that nobody looks at.
  • total is sent as a string in one place and as a number in another.
  • The Flutter app and the web app use different names for the same event.

With trackplan, the plan is the single source of truth and the call becomes:

analytics.checkoutCompleted(
  orderId: 'A12',
  total: 50.0,
  paymentMethod: PaymentMethod.card,
);

Requirements

  • The tool needs Dart SDK 3.11 or later. With Flutter, check the bundled version with flutter --version.
  • Generated code needs Dart language version 3.8 or later in your package.
  • Generated code has no dependencies; trackplan is only a dev dependency.

Quick start

dart pub add dev:trackplan
dart run trackplan init        # writes a sample tracking_plan.yaml
dart run trackplan generate    # writes lib/analytics/analytics.g.dart

import 'analytics/analytics.g.dart';

void main() {
  // Replace DebugAnalyticsAdapter with a real adapter, see "Adapters" below.
  final analytics = Analytics([DebugAnalyticsAdapter(print)]);
  analytics.signUpCompleted(method: SignUpMethod.google);
}

A plan looks like this:

version: 1

targets:
  dart:
    output: lib/analytics/analytics.g.dart

enums:
  payment_method:
    values: [card, paypal, cash]

events:
  checkout_completed:
    description: User finished payment successfully.
    params:
      order_id:
        type: string
      total:
        type: double
        min: 0
      payment_method:
        type: enum
        enum: payment_method
      coupon_code:
        type: string
        required: false
        max_length: 32

Parameter types are string, int, double, bool, enum and list<string>. Event and parameter names are snake_case in the plan and on the wire, and camelCase in Dart. Names that are reserved words get a trailing underscore in Dart (class becomes class_), while the name sent to your backend stays the same.

Commands

Command What it does
dart run trackplan init Writes a sample tracking_plan.yaml. Never overwrites a file.
dart run trackplan validate Checks the plan and prints errors with line and column. --json for tools.
dart run trackplan generate Writes the Dart file.
dart run trackplan check Exits with 1 if the Dart file is missing or out of date. Use it in CI.
dart run trackplan docs Writes TRACKING_PLAN.md, an event catalog for product managers.

Options:

Option Commands
--plan <path> (default tracking_plan.yaml) all
--out <path> generate, check, docs
--allow-outside (allow output paths outside the project) validate, generate, check, docs
--json validate
--help all
--version on its own: dart run trackplan --version

Exit codes: 0 success, 1 the plan has errors or check failed, 2 the tool could not run (bad arguments, missing file).

Example output:

tracking_plan.yaml:18:15 error TP030 Unknown type "float". Did you mean "double"?
1 error, 0 warnings.

Adapters

The generated Analytics class sends every event to a list of adapters. An adapter implements one method:

abstract interface class AnalyticsAdapter {
  void track(String event, Map<String, Object> params);
}

The snippets below were checked against firebase_analytics 13.0.0 and posthog_flutter 5.55.0. Copy them into your app; trackplan itself does not depend on any analytics SDK.

Firebase Analytics

import 'dart:async';

import 'package:firebase_analytics/firebase_analytics.dart';
import 'package:flutter/foundation.dart';

import 'analytics/analytics.g.dart';

/// Sends events to Firebase Analytics.
///
/// Firebase accepts only string and number parameter values, so booleans
/// and lists are converted here.
class FirebaseAnalyticsAdapter implements AnalyticsAdapter {
  FirebaseAnalyticsAdapter([FirebaseAnalytics? analytics])
    : _analytics = analytics ?? FirebaseAnalytics.instance;

  final FirebaseAnalytics _analytics;

  @override
  void track(String event, Map<String, Object> params) {
    unawaited(
      _analytics
          .logEvent(
            name: event,
            parameters: {
              for (final MapEntry(:key, :value) in params.entries)
                key: switch (value) {
                  // Use 1 and 0 instead if you prefer numeric values.
                  bool() => value.toString(),
                  List<String>() => value.join(','),
                  _ => value,
                },
            },
          )
          .catchError((Object e) => debugPrint('Firebase logEvent failed: $e')),
    );
  }
}

Add settings: {warn_provider_limits: [firebase]} to your plan to get warnings for Firebase's documented limits: event and parameter name length, reserved event names and prefixes, number of parameters per event, string length, parameter types other than string and number, and the number of distinct events.

PostHog

import 'dart:async';

import 'package:flutter/foundation.dart';
import 'package:posthog_flutter/posthog_flutter.dart';

import 'analytics/analytics.g.dart';

/// Sends events to PostHog.
class PostHogAnalyticsAdapter implements AnalyticsAdapter {
  @override
  void track(String event, Map<String, Object> params) {
    unawaited(
      Posthog()
          .capture(eventName: event, properties: params)
          .catchError((Object e) => debugPrint('PostHog capture failed: $e')),
    );
  }
}

import 'dart:async';

import 'analytics/analytics.g.dart';

/// Sends events to your own backend in batches.
///
/// [send] is your code, for example an HTTPS POST to your collector. Events
/// are only kept in memory until [flush] is called.
class BatchingAnalyticsAdapter implements AnalyticsAdapter {
  BatchingAnalyticsAdapter(this.send, {this.batchSize = 20});

  final Future<void> Function(List<Map<String, Object>> batch) send;
  final int batchSize;
  final _queue = <Map<String, Object>>[];

  @override
  void track(String event, Map<String, Object> params) {
    _queue.add({'event': event, 'params': params});
    if (_queue.length >= batchSize) {
      // A failed batch stays queued and is retried on the next flush.
      unawaited(flush().catchError((Object _) {}));
    }
  }

  /// Sends all queued events. If [send] throws, the events are queued again.
  Future<void> flush() async {
    if (_queue.isEmpty) return;
    final batch = List.of(_queue);
    _queue.clear();
    try {
      await send(batch);
    } catch (_) {
      _queue.insertAll(0, batch);
      rethrow;
    }
  }
}

/// Drops events until the user has given consent.
///
/// This only filters events sent through trackplan. Analytics SDKs also
/// collect events on their own (for example app opens and sessions), so
/// disable their collection until consent as well, for example with
/// `FirebaseAnalytics.instance.setAnalyticsCollectionEnabled(false)` or
/// PostHog's `optOut()`.
class ConsentAnalyticsAdapter implements AnalyticsAdapter {
  ConsentAnalyticsAdapter(this.inner, {required this.hasConsent});

  final AnalyticsAdapter inner;
  final bool Function() hasConsent;

  @override
  void track(String event, Map<String, Object> params) {
    if (hasConsent()) inner.track(event, params);
  }
}

Debug logging

DebugAnalyticsAdapter passes one line per event to a callback you choose, for example debugPrint. It logs full payloads, which may include user or device identifiers, so only add it in debug builds:

import 'package:flutter/foundation.dart';

final analytics = Analytics([
  FirebaseAnalyticsAdapter(),
  if (kDebugMode) DebugAnalyticsAdapter(debugPrint),
]);

Testing

Every generated file includes RecordingAnalyticsAdapter, which keeps events in memory:

test('checkout sends the order', () {
  final recorder = RecordingAnalyticsAdapter();
  final analytics = Analytics([recorder]);

  analytics.checkoutCompleted(
    orderId: 'A12',
    total: 50,
    paymentMethod: PaymentMethod.card,
  );

  expect(recorder.events.single.event, 'checkout_completed');
  expect(recorder.events.single.params['order_id'], 'A12');
});

min, max and max_length from the plan become asserts, so tests and debug builds fail on out-of-range values. Release builds do not check them.

CI

Fail the build when someone edits the plan but forgets to regenerate:

dart run trackplan check

The generated file contains // dart format off after its header, so dart format in your project leaves it unchanged and check keeps passing.

Library API

The CLI is the main interface. For build tools there is also a small library: validatePlan returns diagnostics and the validated plan, generateDart and generateDocs turn the plan into file contents.

Privacy

trackplan makes no network calls and collects no telemetry. Analytics payloads in your app can contain personal data such as device identifiers. Where the law requires consent, gate both trackplan's adapters (see ConsentAnalyticsAdapter above) and your analytics SDK's own collection, and keep personal data out of event parameters where you can.

License

MIT

Libraries

trackplan
Type-safe analytics code generation from a YAML tracking plan.