apexstock@0.5.1

Full changelog: CHANGELOG.md · npm: https://www.npmjs.com/package/apexstock/v/0.5.1

Changed

• The simple moving average is now the shared one from apex-commons, and
  every number it draws is unchanged. The same window arithmetic sat here and
  in the analysis layer, so it moved to the package both depend on; the OHLC
  accessor, the memo cache and the 2-decimal display rounding stay here, since
  those belong to this product rather than to the average. Verified exact
  rather than assumed: over 360,000 sampled positions (120 price-like series,
  periods 5 to 200) the shared implementation and the one it replaces agreed on
  every one. apex-commons moves from ^0.4.0 to ^0.7.0 and is inlined into
  the bundle as before, which grows by 223 bytes.

  Only the simple average moved. The exponential average and the Bollinger
  bands stay local on purpose and now say why in their own comments: the EMA
  rounds inside its recursion, which the streaming steppers for EMA, MACD and
  TSI reproduce step for step and an exactness test pins, and the bands measure
  their deviation against the rounded middle band that is actually drawn.
  Delegating either would change values the product displays.

• A window average no longer depends on how much history precedes it. This
  was already true and is now a test, because the shared implementation could
  have quietly cost it. Computing a rolling mean with a running total (add the
  new value, subtract the old) is the obvious faster shape and makes the result
  carry the rounding error of everything before it, so the same window averaged
  inside a long series and averaged alone occasionally round to a different
  cent. That property is what lets the appendData() path extend an indicator
  by recomputing only the tail window and still match a full recompute exactly.

• The test harness and the examples now load the ApexCharts a user installs.
  Nothing in the repo declared the engine it renders with: every e2e fixture,
  example, and framework demo pulled apexcharts.js from a sibling checkout on
  the maintainer's disk. A clone plus yarn install could not run the e2e suite
  at all, and when it did run it validated whatever happened to be in that
  directory, which is not what the manifest promises or what a user gets. The
  Angular demo's lockfile shows how far that can drift: it had the engine pinned
  at 5.15.2, two majors behind the peer range declared at the time. ApexCharts
  is now a devDependency pinned to the same range as the peer; fixtures load
  /node_modules/apexcharts/dist/apexcharts.js from a server rooted at the repo
  (it used to be rooted at the parent directory), and examples and demos use the
  relative equivalent. No runtime code changed: the newly installed 7.1.0 is
  byte identical to the checkout the suite had been running against, so the 83
  e2e tests, visual baselines included, pass unchanged.

• The three framework wrappers ship at 0.2.2 with peer ranges that agree with
  each other. react-apexstock, vue-apexstock, and ngx-apexstock had
  gained the apexcharts peer without a release, so npm still served 0.2.1,
  which declared no engine requirement at all. Bumping them surfaced a defect in
  the pairing: apexstock: ">=0.5.0 <1" is the only range that agrees with
  apexcharts: "^7.1.0", because apexstock 0.3.x requires ^5.15.0 and 0.4.x
  requires ^6.7.0. Left as ">=0.3.0 <1" the two peers were mutually
  unsatisfiable, and npm ci (in the publish workflow) and npm install (for a
  user) both failed with ERESOLVE. Their lockfiles were stale for the same
  reason and now resolve apexstock 0.5.0 and apexcharts 7.1.0.

vue-apexstock@0.2.2

Full changelog: CHANGELOG.md · npm: https://www.npmjs.com/package/vue-apexstock/v/0.2.2

The Vue 3 wrapper for ApexStock, published with provenance.

Wrapper releases track the core: see the CHANGELOG for what changed, and packages/vue-apexstock/README.md for the component API.

react-apexstock@0.2.2

Full changelog: CHANGELOG.md · npm: https://www.npmjs.com/package/react-apexstock/v/0.2.2

The React wrapper for ApexStock, published with provenance.

Wrapper releases track the core: see the CHANGELOG for what changed, and packages/react-apexstock/README.md for the component API.

ngx-apexstock@0.2.2

Full changelog: CHANGELOG.md · npm: https://www.npmjs.com/package/ngx-apexstock/v/0.2.2

The Angular wrapper for ApexStock, published with provenance.

Wrapper releases track the core: see the CHANGELOG for what changed, and packages/ngx-apexstock/README.md for the component API.

apexstock@0.5.0

Full changelog: CHANGELOG.md · npm: https://www.npmjs.com/package/apexstock/v/0.5.0

The largest release since the library went commercial: sixteen additions, seven
behaviour changes, fifteen fixes.

One breaking change. The apexcharts peer requirement moves to ^7.1.0
(was ^6.7.0). Upgrade both together:

npm install apexstock@0.5.0 apexcharts@^7.1.0

Two reasons it had to move. The old range had gone stale, and did not admit the
7.x line that every example and test fixture here actually loads, so the suite
was validating a combination the manifest forbade. And the crosshair fix below
lives in ApexCharts 7.1.0: on a zoomed chart with indicators the crosshair drew
on the wrong bar, off by one bar-width per warm-up bar (259px three wheel
notches in, 2056px at nine), and the main chart and each oscillator pane
disagreed with each other because their warm-ups differ.

The headline additions are a financial-analysis workspace (getRangeStats,
getDrawdown, measureRange, a drawdown pane, weighted pane heights),
comparison v2 with baseline policies and a benchmark role, cross-chart
sync, event markers, an on-chart data legend, price-scale modes
(linear / log / percent / indexed), a theme preset pack with
registerTheme, toolbar customization, and a unified export() covering
png / svg / pdf / csv / json, the PDF written without adding a dependency.

Also new is rangeChanging, a per-frame companion to rangeChange. If you have
custom chrome that must follow a zoom gesture, subscribe to it: rangeChange
fires once when the gesture settles, by design, so chrome hung off it alone
freezes for the duration of a wheel zoom.

The three framework wrappers are released alongside at 0.2.2 with the matching
peer requirement.

Added

• Drawdown as a first-class pane, and per-pane heights. Phase 4 of the
  financial-analysis workspace (see plans/financial-analysis-workspace.md):
  price on top, how far underwater it is below, on one shared x-axis and zoom.

  • updateIndicator("drawdown") toggles an analysis pane plotting percent below
    the running peak, with zero pinned to the top of the axis (a drawdown is
    never positive) and the deepest point labelled on the pane itself. It is an
    ordinary oscillator-registry entry, so it inherits the whole pane machinery:
    stacking with other panes, the shared zoom, the theme rebuild, getDataAt,
    getState/setState, and the settings-free dropdown entry. No new pane
    system, and no separate API.
  • It measures on the chart's analysis.drawdownBasis, so the pane,
    getRangeStats(), and getDrawdown() cannot disagree, and the basis
    round-trips in state with the pane.
  • It streams incrementally on appendData: the state is the running peak, so
    the step is O(1) with no warm-up period. The pane's "max" label is
    re-asserted after each append, so a new deeper trough moves it instead of
    leaving a stale line behind.
  • The indicators dropdown groups it under an Analysis heading (a new
    optional group on a registry entry), keeping analysis views distinct from
    technical indicators without adding a second control.
  • Per-pane heights. Panes now divide the indicator area in proportion to a
    relative heightRatio rather than always evenly: the drawdown pane takes 1.4
    shares (it is cumulative, so it earns the room) and everything else takes 1.
    Configure with panes: { drawdown: { heightRatio: 2 } } at construction, or
    setPaneHeightRatio(key, ratio) / getPaneHeightRatios() at runtime (null
    restores the pane's default). The heights always add up to the container
    exactly, and the ratios you set are captured by getState() as a new panes
    key.
    See the README "The drawdown pane" section.

• Comparison persists, and the analysis can be exported. Phase 5 of the
  financial-analysis workspace (see plans/financial-analysis-workspace.md).

  • getState() gains a comparison key: the mode, benchmark, alignment policy,
    and each instrument's name and color. Instrument data is deliberately not
    captured: your app fetches it, it runs to thousands of bars per instrument,
    and it would be stale the moment it was written to storage. So setState()
    restores the setup, keeps any instrument whose data is still loaded, and
    emits the new comparisonRestoreNeeded event with { names } for the
    rest, which the consumer re-supplies with addComparison (the remembered
    color is reapplied). A benchmark whose instrument has not come back yet is
    remembered by name, with the primary filling the role until it does. Same
    division of labour as the price lines' interactive callbacks.
  • Measurements need no state key of their own: a measurement is a measure
    drawing, so it already round-tripped inside drawings.
  • exportData / export gain include, which carries the analysis into an
    export and means something appropriate to each medium. For csv/json:
    "indicators" adds one column per active indicator series (main-chart
    overlays and oscillator panes, null through each warm-up) and
    "analysis" adds return and drawdown per bar. For pdf: "analysis"
    sets a text summary below the chart image (the window's dates and bar count,
    change, high, low, average, volume, annualized return, volatility, max
    drawdown, and the comparison leaderboard when one is active), overridable
    with summary: string[]. The PDF summary follows range, defaulting to the
    visible window there, so the exported numbers describe what you were looking
    at.
  • The OHLC columns are always present, so a CSV keeps round-tripping through
    ApexStock.fromCSV whatever you add, and an extra column whose name collides
    with a spine column is suffixed rather than overwriting it. Range statistics
    are a summary, not a per-bar value, so they are not columns; read them from
    getRangeStats().
  • The PDF summary is still dependency-free: buildPdfFromJpeg grows the page
    for a text band below the image and sets it in Helvetica, one of the PDF
    base-14 fonts, so there is nothing to embed.
    See the README "State Persistence" and "Exporting the analysis" sections.

• Comparison v2: instrument alignment, baseline policies, indexed / relative /
  ratio modes, the benchmark role, and getComparisonStats(). Phase 3 of the
  financial-analysis workspace (see plans/financial-analysis-workspace.md):
  comparison mode now answers "who's up more" fairly even when the instruments
  do not share a calendar.

  • Every instrument, the primary included, is aligned onto one shared x grid
    before anything is normalized, so a newer listing, a different holiday
    calendar, or a missing day no longer skews the comparison.
    setComparisonOptions({ join, fill, baseline, indexBase, source, rebaseRatio, resample }) / getComparisonOptions() expose the policies:
    join ("union" by default, or "primary" to resample onto the primary's
    bars, or "intersection") and fill ("hold" by default, or "gap", or
    "drop").
  • A carried-forward value is used for the math but is not plotted, so a
    line never shows a bar its instrument does not have.
    getComparisonStats() reports how many points were filled.
    resample: true plots the filled grid instead (the default for
    join: "primary", which exists to resample).
  • Baseline policies: "common" (the new default) rebases at the first x
    where every instrument has data; earlier history still plots, as a negative
    percent, rather than being hidden. "own" is the previous behavior, kept for
    compatibility. "visible" rebases to the left edge of the visible window and
    follows the zoom. A number is an explicit x value.
  • Three new modes on top of absolute and percent: indexed (the baseline
    reads indexBase, default 100, the "100 = starting value" view), relative
    (percentChange(asset) - percentChange(benchmark), in percentage points,
    where zero means "kept pace"), and ratio (asset / benchmark, rebased to
    indexBase).
  • The benchmark is a role, not a ticker:
    setComparisonBenchmark(name | "__primary__") /
    getComparisonBenchmark() points it at any added instrument or the chart's
    own symbol (the default). In relative and ratio mode the benchmark's own
    line becomes the flat reference, so what everything is measured against is
    visible on the chart. Removing that instrument hands the role back to the
    primary. No symbol is hard-coded anywhere in the library.
  • getComparisonStats({ from?, to? }) returns the leaderboard ready to render:
    one row per instrument (primary included) with change, excess return vs the
    benchmark, high and low, volatility, worst drawdown, rank, and grid coverage.
    The window runs from the baseline to the last observation, so with
    baseline: "visible" the rows follow the zoom. Volatility and drawdown are
    measured on each instrument's own observations, so a weekly line on a
    daily grid is not annualized as if it had 252 bars a year.
  • New event comparisonChange
    ({ reason, mode, benchmark, baseline, instruments, stats, warnings }),
    carrying the recomputed leaderboard and skipped entirely when nothing is
    subscribed. baseline reports the policy actually applied, which can fall
    back to "own" when no x value has data for every instrument.
  • Declarative defaults via
    analysis: { comparison: { mode, benchmark, join, fill, baseline, indexBase, source } }. A benchmark named there is remembered until its instrument is
    added, with the primary filling the role (and saying so in warnings) until
    then. Unknown option values are warned about and ignored rathe...