FIMViz.jsAPI
    Preparing search index...

    Function parseSciwrid

    • Read a multi-dimensional scientific file into a Dataset with a real temporal axis. This is the implementation behind FimViz.parseFile/fim.addDataset for .nc/.grib2/.zarr — use those; this export exists for the composition root and tests, exactly like parseSource. Every option documented below is passed straight through from them.

      Nothing is decoded here — scan() reads metadata only, and the returned Dataset is a lazy series. Each axis entry is an in-file selector (ref: { select: { variable, time } }), so selecting a timestep costs no second fetch: the child shares this Dataset's bytes and decodes one slice on force. Every op, Stats, ColorScale and RasterLayer then work on it unchanged.

      const ds = await fim.addDataset(file);         // a 120-step NetCDF4 → a time axis
      const t  = ds.select(Date.parse('2023-08-28T06:00:00Z'));   // → one grid, lazily
      await fim.addLayer(t);
      await ds.reduce('mean').grid();                 // temporal mean over the whole axis

      Axis coordinates are numbers, not ISO strings, so select()'s nearest-match works (it is numeric-only) — which is what a time slider needs. Epoch milliseconds (axis.unit === 'ms') when the file carries CF times or the caller supplies dates; plain positions ('index') when the file declares a leading dimension but no labels for it, which is the NetCDF3 case. The original label is kept on each entry's meta.time. This deliberately differs from the WaterML/NWIS adapter's string coords, where exact match was acceptable because latest() covered the common case.

      One variable per Dataset: call it once per variable you want. (Folding variable in as a second axis is roadmapped — §8 — but a variable axis cannot be reduce()d meaningfully, so it needs a guard this first slice does not yet have.)

      Parameters

      • source: string | ArrayBuffer | ArrayBufferView<ArrayBufferLike> | Blob | File | URL
      • Optionalopts: {
            allowExtraDims?: boolean;
            dims?: { order?: "yx" | "xy" };
            grid?: { bbox?: number[]; height?: number; width?: number };
            lon?: "native" | "-180..180" | "0..360";
            name?: string;
            resolveUrl?: (url: string) => string;
            series?:
                | false
                | {
                    coords?: Function
                    | (string | number | Date)[];
                    length?: number;
                    name?: string;
                    unit?: string;
                };
            variable?: string;
            workers?: number;
        } = {}
        • OptionalallowExtraDims?: boolean

          proceed with a variable carrying dimensions beyond (lat, lon) + the series axis — a vertical level, ensemble member or band. Off by default: the reader collapses them with no say from the caller, so this is an acknowledgement, not a fix. Recorded on meta.extraDims

        • Optionaldims?: { order?: "yx" | "xy" }

          which trailing pair of the shape is (lat, lon). 'yx' (CF order, the default) or 'xy' for a variable declared (…, lon, lat). Decides native height/width only — extractGrid resamples onto whatever is requested

        • Optionalgrid?: { bbox?: number[]; height?: number; width?: number }

          PARTIAL override of the native grid; anything omitted comes from the variable's own shape / scan().bbox. Pass bbox alone for a file whose extent scan() could not derive (2-D curvilinear coordinates — ocean tos-style products, rotated poles, Zarr with no CF coords); pass width/height alone to decode coarser than native

        • Optionallon?: "native" | "-180..180" | "0..360"

          re-express the extent in a longitude convention. A global grid is genuinely rolled (the reader cannot do this — asking it for a shifted bbox returns the same pixels relabelled); a regional extent a whole turn away is relabelled with no pixel work; anything else throws rather than splitting the grid

        • Optionalname?: string

          Dataset name; defaults to the filename/URL tail

        • OptionalresolveUrl?: (url: string) => string

          host CORS-proxy/mirror, applied at force time

        • Optionalseries?:
              | false
              | {
                  coords?: Function
                  | (string | number | Date)[];
                  length?: number;
                  name?: string;
                  unit?: string;
              }

          the series (time) axis. Omit for the default: the file's CF times when it has them, otherwise integer indices over its leading dimension. false forces a single grid. coords is an array of one coordinate per step, or a generator (i, n) => coord; ISO strings and Dates become epoch ms, so an unlabelled file gains a REAL time axis — { series: { coords: i => new Date(Date.UTC(2001, i, 1)) } }. length caps the step count (default: the leading dimension), name defaults to 'time', and unit defaults to 'ms' for dates or 'index' for synthesized positions

        • Optionalvariable?: string

          which variable; defaults to the first supported one

        • Optionalworkers?: number

          extractGrid's worker count; defaults to SciWrid's own in a browser and to 0 (inline) under Node, where the worker pool never resolves

      Returns Promise<Dataset>

    Settings
    Member Visibility