Data and archives

CSV/TSV preview#

What it does. Detects the delimiter, parses quoted fields, infers per-column types, reports numeric statistics, and previews the first 100 rows.

How to reach it. Drop a .csv/.tsv (or any file that content-sniffs as delimiter-consistent text - see pipeline.md). Built in csv.js.

How to use it. The preview automatically mounts the full tablekit.js workbench below it (see "Table analysis workbench" below) for interactive sorting/filtering/charting. A file that looks like a Gyroflow IMU log is detected and handed to gcsv.js's gyro-specific view instead.

Table analysis workbench#

What it does. A full data-analysis toolkit mounted below any CSV/XLSX/XLSB/ODS table: a virtualised grid (only visible rows are drawn), column sort/filter/search/hide/reorder/split, a stats bar, group-by summaries, and a chart builder covering line/bar/scatter/histogram/pie/ area/box/heatmap on one hand-drawn canvas (no chart library) with auto-suggested charts and PNG/JSON/CSV export.

How to reach it. Automatic wherever a table renders - csv.js, xlsx.js, xlsb.js, odf.js (ODS) all lazily mount it. Built in tablekit.js.

How to use it.

• Grid: click a column header to sort (ascending, then descending, then back to original order) and select the whole column (shows its stats below - text columns report distinct counts and most-common values); click a row number to select that row; drag across cells to select a rectangle (summarised live in the stats bar).
• Columns menu shows/hides individual columns.
• Split column: pick a column and delimiter (auto-detected by default) to break one packed column into several real columns, splitting its header name too; Undo split reverts. Column Up/Down buttons reorder columns.
• Filter: the gear on each header opens a filter popup (Apply/ Clear) - a min/max range for numbers, a "contains" match for text; the search box filters across every column at once.
• Stats bar: for the current selection, reports count/sum/average/ median/min/max/standard deviation.
• Summarise ("For each ... show the ..."): pick a group-by column, a target number column, and an aggregate (count/sum/average/min/max/ median - count needs no number column) via the group-by key/aggregate/ value selects; rows roll up into per-group totals live. Plot as bar charts the result directly.
• Charts: chart-type buttons pick the type; fields update the drawing instantly; hovering a bar/point/slice shows its exact value; suggested chart buttons are one-click presets; Plot selected range charts whatever's currently selected in the grid; Stacked: on/off toggles stacked mode.
• Export: Export chart PNG, Export stats JSON, Export rows CSV (respects current filter/sort).

Chart and plotting controls (the chart-type buttons are a segmented toggle):

Stacked is an on/off toggle:

Each column filter popup, and the workbench exports:

Notes / limits. Rows are capped (SAMPLE_CAP in table-stats.js) and chart points decimated (MAX_POINTS in tablekit.js) to keep the canvas responsive on large datasets.

Table statistics helpers (shared)#

What it does. Pure, DOM-free math shared by csv.js and tablekit.js: column type inference, numeric description (mean/median/ stddev/percentiles), Pearson correlation, and grouping.

Notes / limits. No controls of its own - a shared library (web/assets/js/lib/table-stats.js). Parses ambiguous D/M-vs-M/D dates day-first, matching the site's European/British convention, rather than Date.parse's US month-first fallback.

Gyroflow IMU log viewer (.gcsv)#

What it does. Parses Gyroflow's generic IMU-log interchange format (the same format Analyser exports from a Sony video's gyro track, see eda-nle.md's Sony rtmd entry): a header of key,value lines (scale factors converting raw columns to real units - time to seconds, gyro to rad/s shown as deg/s, accel to g) followed by a column-header line and samples, plotted via the same gyro/accel timeline viewer sony-rtmd.js uses.

How to reach it. Drop a .gcsv. Built in gcsv.js.

Notes / limits. Capped at 64MB text read; the trace is decimated to 4000 points to keep the canvas responsive.

Structured-data and text-art viewers (HAR / JSON supersets / NFO)#

What it does. Three lightweight viewers sharing only a "show what people actually care about" goal: HAR (.har, an HTTP Archive JSON capture) rendered as a request table (method, status, type, size, time, URL) with a summary; JSON supersets (.json5/.jsonc/.hjson) shown as selectable source, and, when they parse after comment/trailing-comma stripping, as an expandable collapsible value tree; NFO (.nfo, scene-release ASCII art) decoded from CP437 (its native code page) and shown in a monospace block.

How to reach it. Drop a .har/.json5/.jsonc/.hjson/.nfo. Built in dataview.js.

How to use it. JSON tree nodes are <details> elements - click to expand/collapse; children are lazily built only on first expand.

Git object viewer#

What it does. Opens git repository objects with no server and no git binary: loose objects (.git/objects/ab/cdef..., zlib-compressed <type> <size>\0<payload>) inflated via the browser's native DecompressionStream (no library) and parsed by type (commit/tag text, tree entry listing, or a blob handed to the main analyser); pack files (.pack) and pack indexes (.idx) read for header + object count.

How to reach it. Drop a loose git object (detected by content - sniffGitObject(), called from the drop pipeline's magic-byte sniff, since loose objects are extensionless) or a .pack/.idx file. Built in gitobject.js.

How to use it. For a blob object, click Analyse blob content to re-run the git blob's sniffed content through the full analyser.

Email message viewer (.eml/.emlx/.mbox)#

What it does. Parses RFC 822/MIME messages: headers (unfolding continuation lines, decoding RFC 2047 encoded words), walks the MIME tree for a displayable body (preferring text/html sanitised, else text/plain), and lists attachments. Sender authentication (SPF/DKIM/ DMARC) is surfaced from the Authentication-Results header.

How to reach it. Drop a .eml/.emlx (Apple Mail, which prefixes a byte-count line and appends a plist trailer) or .mbox (concatenates many messages, each starting with a From separator line). Built in email.js.

How to use it. For an .mbox with multiple messages, click Show more messages to reveal additional ones.

Notes / limits. No network fetches and no scripts ever run - the HTML body is sanitised the same way the MHTML viewer sanitises saved web pages.

Archive inspection (ZIP and beyond)#

What it does. Inspects ZIP archives by reading the central directory (no full extraction needed for listing), with treemap/breakdown/tree views shared with the folder viewer. Also handles RAR/7z (via lazy libarchive WASM), and raw single-stream compression (gzip, XZ, LZ4, legacy .Z, LZMA) via dedicated decompressors.

How to reach it. Drop a .zip/.rar/.7z/.gz/.xz/.lz4/.Z/ .lzma or other recognised archive extension. Built in archive.js, sharing folder-archive-shared.js for the treemap/breakdown/tree UI.

How to use it. Verify entry CRCs (N files) recomputes each entry's CRC-32 and compares it to the value stored in the archive (integrity check). For a single-file-compressed archive (e.g. a bare .gz), Analyse <inner file> opens the decompressed inner file in the full analyser. The shared Filter by type control (from folder-archive-shared.js) opens a popup of per-extension chips to narrow the treemap.

Notes / limits. ZIP parsing (zip.js) prefers a central-directory read with ranged fetches per entry (order-independent, never loads the whole file), falling back to a sequential local-header walk for truncated/ headerless archives - see renderers.md.

Folder overview#

What it does. Recursively walks a dropped folder (via webkitGetAsEntry) and renders a treemap + summary, sharing the same breakdown/view-toggle UI as the archive viewer.

How to reach it. Drop a folder onto the page. Built in folder.js.

How to use it. Copy paths (N formats) copies one sample file path per unrecognised format found in the folder, useful for reporting a gap; Show folder analysis reveals the full analysis panel; Check which files open (N) scans every file in the folder and flags any that can't actually be opened (a file only counts as "can't open" if its sole viewer path fails outright - unreadable bytes, a HEIC that won't convert, or a RAW with no usable preview or working decode; everything else - metadata, a hex dump, or a "browser can't preview this" banner - counts as openable).

Notes / limits. The openability scan has a 45-second timeout per file so one stuck file can't hang the whole scan.

Treemap visualisation#

What it does. A nested, squarified treemap (WizTree-style): every file in every subfolder rendered at once, leaf rectangles sized by byte size and coloured by category, nested inside folder frames, canvas-based with hover tooltips.

How to reach it. Shared component powering both folder.js's and archive.js's size-breakdown views. Built in treemap.js.

How to use it. Click a tile to zoom into that folder, or open a file's action menu (Cancel/Copy/Analyse); the breadcrumb ("All files" / folder buttons) navigates back up the zoom stack; a tile representing many small files opens an aggregated popup list with a close (×) button, a filter-by-name search box, and a row click per file to analyse it directly.

Comic book archive viewer (CBZ/CBR/CBT/CB7)#

What it does. Mirrors the PDF viewer's UX: a metadata card, a page-thumbnail grid, and a click-to-open lightbox reader with prev/next (and arrow-key) navigation. Pages are image files inside the archive, extracted on demand so a 200-page comic doesn't build hundreds of object URLs up front.

How to reach it. Drop a .cbz (a ZIP), .cbt (a TAR, walked by a minimal built-in reader), or .cbr/.cb7 (via the lazy libarchive unrar/7z WASM loader). Built in comic.js.

How to use it. Click Read (N pages) to open the page reader/ lightbox starting from page 1.

MIDI (Standard MIDI File)#

What it does. A hand-written SMF parser: header, tempo map, time/key signature, track names, General MIDI instruments (program changes), note counts, and duration.

How to reach it. Drop a .mid/.midi. Built in midi.js.

Notes / limits. No playback - MIDI is a score, not sampled audio, and browsers can't decode it directly. No interactive controls, a readout only.

Subtitle files (SRT/WebVTT/ASS/SSA/MicroDVD/SubViewer)#

What it does. Parses cues into a timed list and reports counts, timing, and styling info.

How to reach it. Drop a .srt/.vtt/.ass/.ssa/.sub. Built in subtitles.js.

Notes / limits. The .sub extension is overloaded - it carries text (MicroDVD frame-based or SubViewer time-based) as well as the unrelated binary VobSub bitmap subtitle format, so the module sniffs which one a given .sub actually is before parsing. No interactive controls, a readout only.

LRC timed lyrics#

What it does. Parses .lrc timed-lyric files: ID tags like [ar:Artist] and timestamped lines like [00:12.34]text, including multiple timestamps per line and enhanced word-level <00:12.34> tags.

How to reach it. Drop a .lrc. Built in lrc.js.

Notes / limits. No interactive controls, a readout of metadata and timed lines.

Visual Studio solution manifest (.sln/.slnx)#

What it does. Parses the project list and build configurations out of a Visual Studio solution: classic .sln (a line-based text manifest - format-version line, # Visual Studio <year> comment, one Project(...) = "Name", "Path", "{GUID}" line per project, and Global/GlobalSection blocks) or the newer XML .slnx (a terser <Solution> tree of <Project Path="...">/<Folder>/<Configurations>), both resolved to the same readable summary. Well-known project-type GUIDs (and, for .slnx, project file extensions) are mapped to a friendly language/kind label (C#, C++, Python, Node.js, etc.).

How to reach it. Drop a .sln/.slnx. Built in vssolution.js.

Notes / limits. No interactive controls, a readout only. Common alongside Unity/MonoDevelop projects (Assembly-CSharp).