davide.trentin/KMPCharts
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions
6 commits 1 branch 0 tags 350 KiB
Commit graph

5 commits

Author SHA1 Message Date
Trentin Davide
0e3b79e2da feat(chart): v0.4.0 portfolio hardening 
Ship 15-task plan via kmp-manager group-by-group flow. Library moves
from "prototype" to "senior-engineer-built KMP library" per 3-reviewer
audit (android-reviewer + mobile-performance-reviewer + mobile-architect).

Critical fixes:
- LICENSE MIT (T1) — legal blocker resolved
- explicitApi() Strict + BCV 0.16.3 + klib enabled (T2) — ABI locked
  baseline: chart-realtime.api 161 LOC + chart-realtime.klib.api 211 LOC
- Internal lockdown: CircularBuffer, TieredBuffer, LodDecimator,
  dataVersion, rememberFrameTick (T3)
- kotlinx-datetime 0.6.2 replaces expect/actual currentTimeMs (T4)
  Platform.kt + .android.kt + .ios.kt deleted
- CircularBuffer.writeIndex Int -> Long (T5) — fix 124d overflow @ 200Hz
- Remove !! and @Suppress("UNUSED_EXPRESSION") from commonMain (T6)
- Remove dead config: maxBufferSeconds, xLabelDecimals,
  SignalConfig.label (T7)

Correctness:
- POSITIVE_INFINITY/Float.MAX_VALUE sentinels -> hasData flag (T8)
- Snapshot.withMutableSnapshot + SnapshotApplyConflictException retry
  in push/addSignal/removeSignal (T9)
- dataVersion mutableLongStateOf — Compose-observable, idle = 0
  recompositions (T10) — RenderLoop.kt deleted, targetFps @Deprecated
- Single buffer.snapshot() per signal per frame via SignalEntry
  per-signal scratch arrays (T11) — render-time snapshot cost halved
- M4 binning Tier1/Tier2: firstTs/V + minTs/V + maxTs/V + lastTs/V
  with chronological dedup (T12) — no C7 vertical spike artifacts.
  TIER1/TIER2 CAPACITY x2 -> x4 (TOTAL 77.4k -> 94.8k)
- NaN/Inf/backward-timestamp guards on push (T13) — silent drop,
  per-signal lastPushedTs tracking, KDoc contract

Test gap fill:
- NumberFormat extracted from AxisRenderer to internal object (T14)
  +38 tests pinning rounding/scientific/time-format behavior
- TieredBufferTest +15 tests: tier roll, window-crossing, clear,
  edge cases (T15)
- LodDecimatorTest +10 LTTB-specific branch coverage tests (T15)

Test count: 33 -> 107 (+74).
ABI: 211/272 -> 161/211 LOC (locked, baseline committed).

Deferred to v0.5.0:
- Bump kotlinx-datetime 0.6.2 -> 0.7.x (needs Kotlin 2.1.20+)
- Remove ChartConfig.targetFps (currently @Deprecated, ignored)
- NumberFormat Long overflow @ 1e19 (safe via current routing)
- LTTB upper-bound inclusive vs snapshot half-open semantic mismatch
- Macrobench setup for -30% frame-time verification

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
 2026-05-21 16:27:15 +02:00
Trentin Davide
d6f51cdef2 feat(ios): iOS build pipeline + XCFramework, fix K/Native source bugs 
Modules: chart-realtime (commonMain, iosMain), root, .paul

- chart-realtime/build.gradle.kts: applyDefaultHierarchyTemplate() + XCFramework("ChartRealtime") over iosX64/iosArm64/iosSimulatorArm64, baseName CamelCase
- commonMain RealtimeChartState/CircularBuffer: switch to kotlin.concurrent.Volatile, add kotlin.jvm.JvmName import (OptionalExpectation)
- commonMain AxisRenderer: replace JVM-only String.format with multiplatform formatFixed/formatScientific/pow10 helpers
- iosMain Platform.ios.kt: NSDate() + timeIntervalSinceReferenceDate + 978_307_200s 1970-epoch offset (K/Native binding has no direct timeIntervalSince1970)
- gradle.properties: remove Linux JDK pin (rely on JAVA_HOME)
- .paul: phases plan, STATE+ROADMAP updates

Acceptance: A1 compile 3 iOS targets, A2 linkDebugFrameworkIosSimulatorArm64, A3 iosSimulatorArm64Test (33 tests, 0 failures), A4 assembleChartRealtimeXCFramework (ios-arm64 + ios-arm64_x86_64-simulator slices), A5 Android assembleRelease/publishToMavenLocal --dry-run.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
 2026-05-21 12:45:21 +02:00
dtrentin
73bbd9ddec feat(chart-realtime): v0.3.0 — dirty flag render, targetFps=30, iOS targets, KDoc, README 
- Dirty-flag render: dataVersion counter in RealtimeChartState, skip Canvas draw
  when no new data since last frame (zero GPU work when sensor idle)
- targetFps=30 default in ChartConfig (pass null for max display refresh rate)
- iOS targets: iosX64, iosArm64, iosSimulatorArm64 + Platform.ios.kt actual
- KDoc: RealtimeChart params, dataVersion, ChartConfig.targetFps, LodMode entries,
  AxisLabelMode entries
- README: fixed Maven coords, added LoD/AxisLabels/Material3 sections, updated
  feature list and performance notes
- Tests: RealtimeChartStateTest (17 tests, dataVersion + signal lifecycle)

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
 2026-05-21 11:11:37 +02:00
dtrentin
554b67e630 refactor(buffer): MIN_MAX tier binning, remove MEAN lod mode 
- TieredBuffer: replace MEAN binning with MIN_MAX (min+max per bin).
  Mathematically composable: MIN(bin_mins)=global_MIN, MAX(bin_maxes)=global_MAX.
  Eliminates tier-boundary visual artifacts. TIER1/TIER2 capacity doubled (2 samples/bin).
- LodMode: remove MEAN. Only MIN_MAX and LTTB remain.
  MEAN introduced values never present in raw signal — disinformation for sensor data.
- ChartConfig: default lodMode changed to MIN_MAX.
- Version: 0.4.1-SNAPSHOT → 0.5.0-SNAPSHOT

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
 2026-05-21 08:57:58 +02:00
dtrentin
b9ee5ddfa3 Initial commit: chart-realtime KMP library v0.4.1-SNAPSHOT 
- RealtimeChart composable with Canvas-based rendering
- TieredBuffer: 3-tier ring buffer (5min full-rate, 10min @10Hz, 45min @1Hz) for 1h data support with ~5MB footprint
- LodDecimator: MEAN / MIN_MAX / MIN_MAX(default) / LTTB render-time decimation
- AxisRenderer: X/Y axis with INSIDE/BESIDE/HIDDEN label modes, T0-relative timestamps, scientific notation
- SignalRenderer: clipRect-bounded path rendering with LoD
- ChartTheme: Material3-compatible dynamic theming (light/dark)
- package: dev.dtrentin.chart

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
 2026-05-21 01:48:51 +02:00