Create File Viewer

Create a custom file viewer that is used by the Datagrok file share browser to display files with specific extensions.

Usage

/create-file-viewer [extensions] [package-path]

Instructions

When this skill is invoked, help the user create a custom file viewer for one or more file extensions. Use the /ui skill for building the UI if necessary.

Step 1: Create the viewer function

Add a function to package.ts (or a separate file imported by it) with the correct annotations. The function must:

• Accept a single file input
• Return a view
• Have the fileViewer tag plus a fileViewer-<ext> tag for each supported extension

//name: myFileViewer
//tags: fileViewer, fileViewer-xyz, fileViewer-abc
//input: file file
//output: view v
export function myFileViewer(file: DG.FileInfo) {
  let view = DG.View.create();

  // Read file content and render it
  file.readAsString().then((content) => {
    let host = ui.div([]);
    // Process and display content
    host.innerText = content;
    view.append(host);
  });

  return view;
}

Step 2: Handle file content

Choose the appropriate method to read file content:

• file.readAsString() - for text-based files (returns Promise<string>)
• file.readAsBytes() - for binary files (returns Promise<Uint8Array>)

Example for binary files:

//tags: fileViewer
//input: file file
//output: view v
//meta.fileViewer: mol, sdf, cif
export function structureViewer(file: DG.FileInfo) {
  let view = DG.View.create();
  let host = ui.div([], 'd4-ngl-viewer');

  file
    .readAsBytes()
    .then(bytes => {
      // Process binary content
      let blob = new Blob([bytes]);
      // Render blob...
    });

  view.append(host);
  return view;
}

Key points

• The function runs when a user clicks on a file with a matching extension in the file share browser
• The view is displayed in the file preview panel
• Return the view synchronously; load file content asynchronously inside it
• meta.fileViewer can take multiple extensions

For extensions that are efficiently DataFrames

• Return a DG.TableView directly — do not wrap a grid in a custom DG.View
• Transfer all source metadata to tags (table-level and column-level) so it survives round-trips through the platform

// Good — file viewer for a single-table format
static async previewFoo(file: DG.FileInfo): Promise<DG.View> {
  const bytes = await file.readAsBytes();
  const data = await parseFoo(bytes);
  const df = toDataFrame(data);
  const view = DG.TableView.create(df, false);
  view.name = file.name;
  return view;
}

• Set df.setTag('source.format', '...') to record where the data came from
• Preserve column-level metadata (description, format, categories) as tags
• Use a namespace prefix for format-specific tags (e.g., minitab.version, prism.sheetId)
• Use the standard description tag for column descriptions

For extensions that contain multiple dataframes

• Use custom DG.View
• Show global file metadata on top
• 'Open all' big button on the ribbon
• Tab control with pane per dataframe

Behavior

1. Ask the user which file extension(s) they want to support if not specified
2. Determine whether the file is text-based or binary to choose readAsString vs readAsBytes
3. Create the viewer function with correct tags for all specified extensions
4. Add rendering logic appropriate for the file type