BinView

A hex editor for VS Code that understands raw binaries and Intel HEX, with a side panel that stacks a whole-file image map over a data inspector.

Features

Hex editor

• Address, hex octet and text columns, with column grouping and 4–64 bytes per row.
• Text column can render ASCII, Latin-1 or CP437 (the classic DOS glyph set).
• Edit directly in either column: type hex digits over the octets, or type characters over the text. Both write through VS Code's normal dirty/save/undo machinery (Ctrl+Z, Ctrl+S, revert and hot-exit all work).
• Paste hex (DE AD BE EF, 0xDE,0xAD) or text; copy the selection as hex, a C array, base64 or text.
• Find with a widget modelled on the editor's own: Ctrl+F, live match counting (3 of 17), every match highlighted, Enter / Shift+Enter to step and wrap, Escape or a click outside to close. A query is read both ways at once - dead finds the bytes DE AD and the word "dead" - and text matching ignores case. Aa demands an exact case match, 0x drops the text reading, and opening find with a short selection seeds it as a byte search. Hover the match count to see the split ("2 as bytes, 1 as text").
• Go to an offset (Ctrl+G), select with mouse or keyboard.
• Hand the selection to CyberChef in one click, from the chef-hat button at the right of the toolbar (see below).

Editing is overwrite only — the file never changes length, which is what you want for firmware images and fixed layouts.

Intel HEX

.hex / .ihex / .ihx / .mcs / .h86 files (and any file whose first bytes look like Intel HEX records) are parsed into address-tagged regions:

• Real addresses in the address column, including extended linear/segment addressing.
• Unmapped space between regions is drawn as a gap of N bytes separator rather than silently concatenated.
• Checksums are verified on load; a bad record is reported instead of being guessed at.
• Saving rewrites the records, preserving the record length, line endings and start address record of the original file, and never straddling a 64 KiB page boundary.

Hiding unwritten space

Firmware images are mostly erased flash. Toggle Hide fill rows in the Config menu (or BinView: Toggle Hiding of Fill Rows) and every run of two or more identical rows collapses into a single ⋯ 1,234 rows of 0xFF hidden ⋯ marker.

Any repeated byte counts, and each run is labelled with its own value, so a file padded with 0x00 in one place and erased to 0xFF in another collapses both without being told which is which. A lone fill row is left visible - it takes a run to be worth hiding - and binview.fillRunMinRows sets how long a run has to be (default 2).

Click a marker to expand that one run. Jumping to an address or finding a match inside hidden space turns the filter off so you can see where you landed. Rows are re-evaluated when you toggle the filter, not while you type, so the row under your cursor never disappears mid-edit.

Data inspector

The lower pane decodes the bytes at the cursor: int/uint 8–64, float32/float64 in both endiannesses, binary, octal, character, the UTF-8 sequence, and the address. The eight bit buttons toggle bits in place — clicking one is an edit.

File map

The upper pane of the side panel renders the whole file as an image and follows the cursor. Click anywhere on it to jump the editor there. Modes:

Hovering reads out the offset and the value under the pointer, phrased for the mode: 0x0004C0 · 0x7F (127) for byte value, · entropy 0.853 for entropy, · ascii for byte class, and 0x41 → 0x42 · 1,024 pairs for the bigram plot. The ? button explains what each mode shows.

Entropy is normalised to 0-1 (1 being the full 8 bits per byte), the way binary analysis tools plot it. In entropy mode a Cut slider sets a threshold on that scale: left at zero the map is the usual continuous ramp; move it and the map goes two-tone - at or above the cut in yellow, below it in dark - which makes the boundary of a compressed or encrypted region obvious. The hover readout keeps reporting the real value either way.

Ctrl+scroll (or a pinch gesture) over the map changes how wide the image is drawn, trading detail against coverage. Large files are folded so several bytes share a pixel - the note under the image says how many - and the image is fitted to the height the pane actually has. The current selection is boxed on the image.

Commands

Inside the grid: arrows/PageUp/PageDown/Home/End move, Shift extends the selection, Tab switches between the hex and text columns, Ctrl+A selects all, Delete zeroes the selection, Backspace steps back.

With a range selected, an unshifted arrow slides the whole selection instead of collapsing it: left/right by the selection's own size, up/down by one row (same columns, next row). So selecting a 4-byte field and holding Down walks that field through a table of records, and holding Right walks it through consecutive fields. The inspector always decodes from the first selected byte, so extending a range never changes what it shows.

Settings

Developing

npm install
npm run build      # bundle into dist/
npm run watch      # rebuild on change
npm test           # format, render and webview (jsdom) tests
npm run package    # produce a .vsix

Press F5 in VS Code to launch an Extension Development Host (or install the packaged build with code --install-extension binview-0.1.0.vsix), then open examples/sample.bin (firmware-shaped: header, code, padding, strings, erased flash) or examples/sample.hex (two regions 256 KiB apart). Regenerate them with node examples/make-samples.mjs.

Layout

Known limits

• Editing is overwrite only: no insert or delete.
• Intel HEX editing changes bytes inside existing records; it cannot add addresses that the file does not already map.
• Matches for search do not span a gap between two Intel HEX regions.

License

MIT