Add feature for Embedded Workflows for Images (PNG, SVG, JPEG)#376
Add feature for Embedded Workflows for Images (PNG, SVG, JPEG)#376GREENRAT-K405 wants to merge 4 commits intoControlCore-Project:devfrom
Conversation
There was a problem hiding this comment.
Pull request overview
Adds “embedded workflow” image export/import so PNG/SVG/JPEG files can carry GraphML metadata and be re-opened in the editor as editable workflows.
Changes:
- Adds new Export menu options for embedded/non-embedded PNG, SVG, and JPEG.
- Implements embedding/extraction of GraphML in PNG (tEXt), SVG (
<metadata>), and JPEG (COM segments with splitting). - Extends drag-and-drop file opening to accept common image formats; adds dependencies for PNG chunk handling and SVG export.
Reviewed changes
Copilot reviewed 6 out of 7 changed files in this pull request and generated 5 comments.
Show a summary per file
| File | Description |
|---|---|
| src/toolbarActions/toolbarList.js | Adds export menu items for embedded/non-embedded PNG/SVG/JPEG. |
| src/toolbarActions/toolbarFunctions.js | Extends readFile to detect and import embedded GraphML from PNG/SVG/JPEG. |
| src/graph-builder/graph-core/5-load-save.js | Implements embedded export logic for PNG/SVG/JPEG, including JPEG segment splitting. |
| src/graph-builder/graph-core/1-core.js | Registers cytoscape-svg extension to enable cy.svg(). |
| src/component/File-drag-drop.jsx | Allows drag/drop and file input selection for image formats. |
| package.json | Adds dependencies for SVG export and PNG chunk encoding/decoding. |
| package-lock.json | Locks new dependency tree for the added libraries. |
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
| "cytoscape-grid-guide": "^2.3.2", | ||
| "cytoscape-node-editing": "^4.0.0", | ||
| "cytoscape-svg": "^0.4.0", | ||
| "file-saver": "^2.0.5", | ||
| "hotkeys-js": "^3.8.7", | ||
| "jquery": "^3.0.0", | ||
| "konva": "^7.0.3", | ||
| "lucide-react": "^0.487.0", | ||
| "md5": "^2.3.0", | ||
| "moment": "^2.29.4", | ||
| "png-chunk-text": "^1.0.0", | ||
| "png-chunks-encode": "^1.0.0", | ||
| "png-chunks-extract": "^1.0.0", |
There was a problem hiding this comment.
New dependency cytoscape-svg is GPLv3-licensed (see package-lock.json entry), while this repo is BSD-3-Clause (LICENSE). This can create a license incompatibility for distribution. Please confirm the project can legally take a GPLv3 dependency or switch to an SVG export approach/library with a permissive license compatible with BSD.
There was a problem hiding this comment.
@pradeeban Actually, excellent catch by copilot there.
I think the main problem is not cytoscape or even svg as entirety but just this specific plugin cytoscape-svg that I used to add the functionality for svg import export (svg support didn't exist prior to this PR).
cystoscape-svg is GPLv3 licensed and will not be ideal for open source, (we should only stick to MIT licensed and open source).
It's an excellent note to take for me as a contributor to verify the licensing for any new packages we add for some new functionality or some fix. Will keep this in mind for future!
I have fixed this with my first follow-up commit.
| let graphMLData = ''; | ||
| while (pos < buffer.length) { | ||
| if (buffer[pos] !== 0xFF) break; | ||
| const marker = buffer[pos + 1]; | ||
| if (marker === 0xDA) break; // SOS - Start of Scan | ||
| const len = (buffer[pos + 2] * 256) + buffer[pos + 3]; | ||
| if (marker === 0xFE) { // COM Comment segment | ||
| graphMLData += new TextDecoder().decode(buffer.slice(pos + 4, pos + 2 + len)); | ||
| } | ||
| pos += 2 + len; | ||
| } | ||
|
|
There was a problem hiding this comment.
JPEG embedded GraphML parsing concatenates TextDecoder().decode(...) results per COM segment. Since the exporter splits the UTF-8 byte stream into 65,533-byte chunks, it can split a multi-byte UTF-8 sequence across segments; decoding each segment independently can corrupt characters (U+FFFD) and make the reconstructed GraphML invalid. Consider collecting COM segment bytes first and decoding once, or use TextDecoder.decode(..., { stream: true }) across segments before a final flush.
| let graphMLData = ''; | |
| while (pos < buffer.length) { | |
| if (buffer[pos] !== 0xFF) break; | |
| const marker = buffer[pos + 1]; | |
| if (marker === 0xDA) break; // SOS - Start of Scan | |
| const len = (buffer[pos + 2] * 256) + buffer[pos + 3]; | |
| if (marker === 0xFE) { // COM Comment segment | |
| graphMLData += new TextDecoder().decode(buffer.slice(pos + 4, pos + 2 + len)); | |
| } | |
| pos += 2 + len; | |
| } | |
| const comSegments = []; | |
| while (pos < buffer.length) { | |
| if (buffer[pos] !== 0xFF) break; | |
| const marker = buffer[pos + 1]; | |
| if (marker === 0xDA) break; // SOS - Start of Scan | |
| const len = (buffer[pos + 2] * 256) + buffer[pos + 3]; | |
| if (marker === 0xFE) { // COM Comment segment | |
| comSegments.push(buffer.slice(pos + 4, pos + 2 + len)); | |
| } | |
| pos += 2 + len; | |
| } | |
| let graphMLData = ''; | |
| if (comSegments.length > 0) { | |
| const totalLength = comSegments.reduce((sum, seg) => sum + seg.length, 0); | |
| const allBytes = new Uint8Array(totalLength); | |
| let offset = 0; | |
| for (const seg of comSegments) { | |
| allBytes.set(seg, offset); | |
| offset += seg.length; | |
| } | |
| graphMLData = new TextDecoder().decode(allBytes); | |
| } |
There was a problem hiding this comment.
fixed with second follow-up commit!
| const b64Uri = this.cy.png({ full: true }); | ||
| const b64Data = b64Uri.split(',')[1]; | ||
| const buffer = new Uint8Array(window.atob(b64Data).split('').map((c) => c.charCodeAt(0))); | ||
| const chunks = extractChunks(buffer); | ||
| chunks.splice(-1, 0, textChunk.encode('graphml', this.getGraphML())); | ||
| const newBuffer = new Uint8Array(encodeChunks(chunks)); |
There was a problem hiding this comment.
PNG-EMBEDDED export converts a base64 data URL to bytes via window.atob(...).split('').map(...), which creates large intermediate arrays/strings and can blow up memory for larger graphs. Prefer a single-pass conversion (e.g., preallocate a Uint8Array and fill it in a loop, or convert via fetch(dataUrl).then(r => r.arrayBuffer())) to reduce peak memory and improve performance.
There was a problem hiding this comment.
fixed with second follow-up commit!
| const b64Uri = this.cy.jpg({ full: true }); | ||
| const b64Data = b64Uri.split(',')[1]; | ||
| const buffer = new Uint8Array(window.atob(b64Data).split('').map((c) => c.charCodeAt(0))); | ||
| const graphMLStr = this.getGraphML(); |
There was a problem hiding this comment.
JPG-EMBEDDED export also converts the base64 data URL to bytes via window.atob(...).split('').map(...), which is memory-heavy for large images. Please switch to a more memory-efficient conversion (single-pass loop into a preallocated Uint8Array, or fetch(dataUrl) to arrayBuffer()).
There was a problem hiding this comment.
fixed with second follow-up commit!
| const ext = droppedFile && droppedFile.name.split('.').slice(-1)[0]?.toLowerCase(); | ||
| const allowed = ['graphml', 'json', 'png', 'svg', 'jpg', 'jpeg']; | ||
| if (e.dataTransfer.files.length === 1 && allowed.includes(ext)) { | ||
| readFile(superStateRef.current, dispatcherRef.current, droppedFile); |
There was a problem hiding this comment.
PR description says the File Open menu scans image files for embedded metadata, but the file picker in src/component/fileBrowser.jsx still restricts selection to .graphml/.json only (both pickerOpts.types.accept and the fallback <input accept>). Please update the File Open menu accept/types to include .png, .svg, .jpg, .jpeg so the UI matches the new readFile support.
There was a problem hiding this comment.
fixed in third follow up commit!
Recording.2026-03-28.175136.mp4@pradeeban please have a look! |
|
@pradeeban will create a PR with updates that are necessary to build study with these embedded type workflows in concore repo |
|
@GREENRAT-K405 can you check the bot comments above? |
Sure @pradeeban, I will get it done shortly! |
There was a problem hiding this comment.
Pull request overview
Copilot reviewed 6 out of 7 changed files in this pull request and generated 3 comments.
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
| let graphMLData = ''; | ||
| if (comSegments.length > 0) { | ||
| const totalLength = comSegments.reduce((sum, seg) => sum + seg.length, 0); | ||
| const allBytes = new Uint8Array(totalLength); | ||
| let offset = 0; | ||
| for (let i = 0; i < comSegments.length; i += 1) { | ||
| allBytes.set(comSegments[i], offset); | ||
| offset += comSegments[i].length; | ||
| } | ||
| graphMLData = new TextDecoder().decode(allBytes); |
There was a problem hiding this comment.
JPEG import concatenates all COM (0xFFFE) segments found before SOS and treats the result as GraphML. This will break for JPEGs that contain unrelated comment segments (or other embedded metadata added by external tools). Consider prefixing your GraphML payload with a unique signature (and optional segment index/count) and filtering/assembling only those segments that match, ignoring unrelated COM segments.
There was a problem hiding this comment.
This is probably too much edge case thinking, the images should not break in normal sharing
No changes required here in my opinion, still will like to know Mentors suggestion!
| const comSegments = []; | ||
| for (let i = 0; i < graphMLBytes.length; i += 65533) { | ||
| const chunk = graphMLBytes.slice(i, i + 65533); | ||
| const segmentLen = chunk.length + 2; | ||
| const seg = new Uint8Array(4 + chunk.length); | ||
| seg[0] = 0xFF; | ||
| seg[1] = 0xFE; | ||
| seg[2] = Math.floor(segmentLen / 256); | ||
| seg[3] = segmentLen % 256; | ||
| seg.set(chunk, 4); |
There was a problem hiding this comment.
JPEG embedded export writes raw GraphML bytes directly into COM segments without any identifier/signature. Since the import path currently reads all COM segments, any pre-existing COM segments will corrupt the payload (and external tools may add COM segments later). Add a recognizable header (and segment index/count) to each COM segment so import can reliably select and reassemble only your GraphML segments.
| const comSegments = []; | |
| for (let i = 0; i < graphMLBytes.length; i += 65533) { | |
| const chunk = graphMLBytes.slice(i, i + 65533); | |
| const segmentLen = chunk.length + 2; | |
| const seg = new Uint8Array(4 + chunk.length); | |
| seg[0] = 0xFF; | |
| seg[1] = 0xFE; | |
| seg[2] = Math.floor(segmentLen / 256); | |
| seg[3] = segmentLen % 256; | |
| seg.set(chunk, 4); | |
| // Prepare a recognizable header for each COM segment carrying GraphML. | |
| const headerBytes = new TextEncoder().encode('GRAPHMLCHUNK'); | |
| const perSegmentMetaLength = 4; // 2 bytes: total segment count, 2 bytes: segment index | |
| const headerLength = headerBytes.length + perSegmentMetaLength; | |
| // JPEG COM segment payload limit is 65533 bytes (length field is 2 bytes, max 65535). | |
| const maxChunkSize = 65533 - headerLength; | |
| const segmentCount = Math.ceil(graphMLBytes.length / maxChunkSize); | |
| const comSegments = []; | |
| for (let segIndex = 0, offset = 0; offset < graphMLBytes.length; segIndex += 1, offset += maxChunkSize) { | |
| const chunk = graphMLBytes.slice(offset, offset + maxChunkSize); | |
| const payloadLength = headerLength + chunk.length; | |
| const segmentLen = payloadLength + 2; // includes these 2 bytes for the length field | |
| const seg = new Uint8Array(4 + payloadLength); | |
| // JPEG COM marker | |
| seg[0] = 0xFF; | |
| seg[1] = 0xFE; | |
| // Big-endian segment length (includes these 2 bytes, excludes marker) | |
| seg[2] = Math.floor(segmentLen / 256); | |
| seg[3] = segmentLen % 256; | |
| // Header: magic string | |
| let writePos = 4; | |
| seg.set(headerBytes, writePos); | |
| writePos += headerBytes.length; | |
| // Header: total segment count (2 bytes, big-endian) | |
| seg[writePos] = Math.floor(segmentCount / 256); | |
| seg[writePos + 1] = segmentCount % 256; | |
| writePos += 2; | |
| // Header: current segment index (0-based, 2 bytes, big-endian) | |
| seg[writePos] = Math.floor(segIndex / 256); | |
| seg[writePos + 1] = segIndex % 256; | |
| writePos += 2; | |
| // Payload: chunk of GraphML bytes | |
| seg.set(chunk, writePos); |
There was a problem hiding this comment.
This is probably too much edge case thinking, the images should not break in normal sharing
No changes required here in my opinion, still will like to know Mentors suggestion!
|
@pradeeban I have taken some improvements from the bot as you had suggested, this PR should be ready for your review, Thanks :) |
feature to let the users export their graphs as standard images (PNG, SVG, JPEG) while embedding the underlying graphml logic directly within the file metadata. These images can be dragged back into concore-editor to instantly reconstruct the interactive graph, making workflow sharing and documentation frictionless.
This closes #375