Client-side CSV block model analyzer for mining and geostatistics. Drop a file, get instant statistics — nothing leaves your browser.

Live app

BMA analyzes block model CSV exports from mining software (Isatis, Vulcan, Surpac, Leapfrog, Datamine). It streams files row-by-row using Web Workers, so multi-gigabyte models work on modest hardware without uploading anything to a server.

• Preflight — auto-detects delimiters, column types, and XYZ coordinates; lets you override types, skip columns, and set per-column value filters before analysis
• Aux — load a second dataset (e.g. the composites/samples behind the model) and compare it against the block model across Statistics, CDF, and Swath, with its own filter, calculated columns, and declustering weights
• Summary — file metadata, grid geometry, block dimensions, sub-block detection, fill ratio, coordinate ordering
• Calc — calculated columns with an expression editor (r.Fe / r.SiO2), live preview, autocomplete, and chained evaluation; a Model | Aux toggle edits either dataset
• Statistics — Welford online stats (mean, std, CV%, skewness, kurtosis), t-digest quantiles, CDF plots, optional row weights (e.g. ore proportion), model-vs-samples comparison rows
• Categories — value counts, frequency bars, Shannon entropy, search and collapse
• StatsCat — statistics grouped by a categorical variable with per-group tables and CDF overlays, cross-tabulation mode
• GT — grade-tonnage curves with per-variable units, density (column or constant), group-by overlays, and SVG/PNG export
• Swath — directional trend plots (X/Y/Z or custom azimuth/plunge), multi-variable overlay or split layouts, model-vs-samples overlay, SVG/PNG export
• Export — stream-export selected/renamed/reordered columns as CSV with drag-and-drop column ordering

Global row filter expressions use JavaScript syntax on row objects: r.LITO === 'BIF' && r.fe_pct > 30. Filters can reference calculated columns.

Open index.html in a browser. No server, no install, no dependencies.

Supports .csv, .txt, .dat, and .zip files. Also installable as a PWA.

The distributable is a single index.html — CSS, HTML, Web Worker (as a template literal), and main-thread JS in one file. No frameworks, no dependencies. Source lives in src/ and is assembled by node build.js (plain concatenation, not a bundler); node check.js syntax-checks the modules and the built output.

Key design decisions:

• Streaming — File.stream() + TextDecoderStream, never loads the full file into memory
• Single-pass — type detection, stats, geometry, and calculated columns all computed in one file read
• Web Worker — all analysis runs off the main thread; a separate worker handles export
• Expression DSL — filters and calcols compile to JS via new Function() with a math preamble (abs, sqrt, clamp, ifnull, between, remap, etc.)

See docs/ for detailed architecture docs.

MIT — Arthur Endlein, 2026