Skip to main content

@tidal-ts/core

The part of tidal-ts with no drawing in it: what a study is, how data arrives, and how time is cut into bars and sessions. Peers: pond-ts, @pond-ts/financial and @pond-ts/process (pinned exactly).

Studies​

A study is described by a plain object, a DeriveSpec:

const macd: DeriveSpec = {
  op: 'macd',
  inputs: ['close'],
  params: { fastPeriod: 12, slowPeriod: 26, signalPeriod: 9 },
};

The same spec always produces the same column name (deriveId), so a study can be saved, shared and recomputed without ambiguity.

The studies come from @pond-ts/financial and run through @pond-ts/process. The study catalog shows every one with a live chart. To read the list in code:

FunctionAnswers
studyCatalog()Every study on offer: its op, display label, one-line summary and family.
opParams(op)Its settings, with defaults and allowed ranges.
opOutputs(op)The lines it produces (one, or several such as Upper / Lower).
opNeedsColumns(op)Which bar columns it reads (empty for a study of one column, such as RSI).
opSharesSourceAxis(op)Whether it is drawn over price, or in a panel of its own.
studyLevels(op)The guide lines it declares, such as 30 and 70 for RSI.
isValidSpec(spec)Whether a spec's settings are legal before you draw it.
applyDerived(series, specs)Compute studies onto a series yourself, without the chart.

Data in​

  • snapshotToTimeSeries(snapshot) turns a column-by-column payload (a name, a row count, a schema and one array per column) into a pond TimeSeries, and appendToRows turns a later update in the same layout into rows to append.
  • buildPriceSeries(name, rows) builds an open/high/low/close/volume series from plain rows.
  • generatePriceSeries and generateVolSeries make seeded practice data for tests, stories and these docs.

Volatility​

VOL_SCHEMA is the layout for a volatility history. Per tenor it holds two at-the-money implied vol curves (iv and hv, which treat earnings days differently; despite the letter, hv is implied, not historical), realized vol, and skew. impliedCol, hvCol, rvCol, rvCenCol and skewCol name a column for a tenor, so code never spells column names by hand.

Bars, sessions and time zones​

  • ohlcWindow rolls bars up into coarser ones (minutes into hours, days into weeks), and autoBarWindow picks the finest bar size that still leaves each candle at least a few pixels wide, so up and down colours stay honest when zoomed far out.
  • sessionSegments and exactSessionSegments find the trading sessions in the data, so the time axis can skip nights and weekends instead of drawing them as empty space. sessionOpenLine starts each session's price line at its open.
  • DATA_TIME_ZONE, TIME_ZONE_CHOICES and resolveTimeZone keep the zone the data is stored in separate from the zone the axis is read in.
  • holdAcrossGrid(series, grid, grainMs) puts a coarse series on a finer grid: a daily curve across one-minute bars. Each row holds from its key for grainMs, or until the next row starts. A grid point never reads a bar that starts after it, and a day with no row stays a gap. Every column comes through, as optional. Hold after computing studies, not before: a 20-bar average of a held daily series is a 20-minute window. The chart's prepare step does this for you with hold (see the chart page).