Data Workspace: Shared Variables, CSV Import, Scenarios

The data workspace is one named-variable table that every CalcX panel shares. Import a CSV, TSV or pasted block once, and the same x or y is available by name to the statistics, curve-fitting, filter, matrix, Python and R panels; anything another panel sends from its Send to menu to Workspace lands here as a variable too. Each variable keeps its shape — scalar, vector, matrix, table, string or boolean — plus an optional provenance note such as zscore(y), and everything is stored in this browser with nothing uploaded.

Updated 2026-10-01Reviewed by CalcX Editorial Team5 sources5 min read

How to use it

Import once.The Import tab takes pasted text or a .csv/.tsv/.txt file and auto-detects the delimiter (comma, tab, semicolon or runs of spaces) and whether the first row is a header. Import as matrix writes one variable; Import by column writes one vector per column under the cleaned header names (my.col becomes my_col).
Inspect and edit.Click a row for its shape and preview, summary cards (n, min, max, mean, sd, non-finite), an editable grid capped at the first 100 rows × 26 columns, and a chart: line or histogram for vectors, a scatter for a two-column matrix. Tables render read-only.
Derive a new variable.Transpose, Normalise 0–1, z-score, Sort, Diff, Cumsum and Take column write a new variable named base_suffix (y_z) with the operation recorded as its provenance note.
Snapshot with scenarios.The Scenarios tab saves a named snapshot of every variable, up to 20. Load replaces the whole workspace after a confirmation, Merge overwrites same-named variables only, and any two snapshots diff by name.
Hand data on.Export the selected variable as CSV, the whole workspace as JSON (values plus notes), or copy MATLAB, Python and R literals — including an init block for every variable — from the Language handoff tab.

Worked example

Paste this into the Import tab:

Step	On-screen readout
After paste	`Delimiter: , · header: x \| y · 3 rows × 2 cols`
Import by column	`Imported 2 variables`; `x` = 1 2 3, `y` = 2.5 4.1 6.7 (both 3-vectors)
Inspect `y`	n 3 · min 2.5 · max 6.7 · mean 4.43333 · sd 2.11975 · non-finite 0
Press z-score	`Created y_z = zscore(y)`; mean -1.4803e-16, sd 1, preview -0.912058 -0.157251 1.06931

The z-score uses the sample sd above, so the mean card of y_z reads -1.4803e-16 — binary arithmetic just off zero, not a typo — while its sd card reads 1. The Language handoff tab prints the same numbers in the target syntax:

Scenarios make the what-if version of the same idea: with scale = 3 saved as Baseline and scale = 5 saved as After, the diff table shows Δ = 2.

text

x,y
1,2.5
2,4.1
3,6.7

x <- c(1, 2, 3)
y <- c(2.5, 4.1, 6.7)
y_z <- c(-0.9120579223975365, -0.15725136593061, 1.069309288328146)

Live values, snapshots and references

The workspace is the application's only live data: editing a variable changes it in place, and a handoff carrying a reference resolves that name when the receiving panel consumes it, so the numbers are not copied. A scenario is the opposite — saving deep-copies every variable, later edits never touch it, and Load or Merge writes the snapshot back; re-save to refresh it. Renaming or deleting a variable breaks references to the old name.

Limits to keep in mind

Grids are capped, the data is not.: The editor shows the first 100 rows × 26 columns and tables are read-only, but the variable keeps every cell and the CSV export writes all of them.
Names are identifiers.: [A-Za-z_][A-Za-z0-9_]* only, no dots, and reserved words such as if, end, pi and ans are refused.
It stores and reshapes, it does not compute.: There is no expression box, model fit or script runner here; use the statistics workbench, the matrix tools, Engineering Lab, the Python or R kernels, or the Compute IDE.

The 20,000-element boundary

The workspace saves variables to local storage as they change, but only up to 20,000 elements. A 200×100 matrix sits exactly on that line and survives a reload; a 201×100 matrix (20,100 elements) stays in memory only — inspect, derive and export it while the tab is open, but after a reload it is gone from the table with no warning. In a direct test of the store, saving small = [1, 2, 3] and then the 201×100 matrix left the saved record holding only small, and a fresh load returned only small. Scenario snapshots filter at the same boundary. Tall imports are safer split by column, because each 201-element vector is far below the line.

Where it fits

One import, several panels

Load an x, y pair once; the statistics, regression, matrix and graphing panels resolve the values by name instead of holding a stale copy.

Baseline versus after

Save a scenario before an experiment and another after; Load restores the baseline, Merge keeps variables the snapshot does not mention, and the diff table gives the per-scalar Δ.

Data into code

Copy the MATLAB, Python or R init block into a script's first line; the MATLAB kernel seeds from the workspace and the R panel can inject it.

References

IETF, RFC 4180: Common Format and MIME Type for Comma-Separated Values (CSV) Files, rfc-editor.org （访问日期：2026-10-01）— delimiter, header and record rules behind the import parser.
NIST/SEMATECH, e-Handbook of Statistical Methods, §1.3.5.2 Mean, Variance, Standard Deviation, itl.nist.gov （访问日期：2026-10-01）— the n − 1 sample sd shown in the inspector.
MDN Web Docs, Window.localStorage （访问日期：2026-10-01）— the per-origin store that keeps variables between reloads.
IETF, RFC 8259: The JavaScript Object Notation (JSON) Data Interchange Format, rfc-editor.org （访问日期：2026-10-01）— the format of the workspace export.
R Core Team, data.frame: Data Frames, stat.ethz.ch （访问日期：2026-10-01）— the R literal emitted for table variables.