Photoshop psd support

[mnaoumov]

It would be handy to include more formats like psd. It seems it is not too difficult to convert it to jpg during import: command line - How to convert PSD to JPG or PNG on Linux with Gimp CLI/batch mode (or other server-side tool)? - Super User

[mrm]

ImageMagick has a pretty horrible security track record, but if someone doesn’t mind the security implications, installs it externally, I guess I could fork out a call to convert and if it returns something, I import it?

The referenced npm module is abandoned, but it’s just a wrapper around GitHub - meltingice/psd.js: A Photoshop PSD file parser for NodeJS and browsers

[mnaoumov]

I haven’t tried it, but the library on your link says it has no native module dependencies. So your ImageMagick security concern could be avoided

[mrm]

ghostscript generates rastered images of .ai and .pdf:

gs -dSAFER -dBATCH -dNOPAUSE -sDEVICE=jpeg -SOutputFile=output.jpeg input.pdf

[sandeep]

So facing this right now. There are two issues: 1. Storage 2. Display.

Storage:
I have PSDs that I would like stored along side the other files they correspond to. So that we have one … well… Structure for Photos… This feels necessary for me.

Display:
The solution above with imgmagick works. But in the interim - I can live with a placeholder image when I drill down into a view and the image not showing up in the summary/random seed based summary views.

This could work for any other filetypes or “related data” that you might want to store with the associated pictures.

[mrm]

Ideally there would be an way for advanced users to “teach” PhotoStructure how to handle file types that it currently can’t handle.

PhotoStructure needs a couple things to import files:

1. A way to extract the “captured-at” time. This can revert to the mtime of the file, but that’s not great.

2. A way to render the file as an image, so it can be presented in the tag galleries and asset views. This could be something as basic as a file icon with the extension, like this:

   
   , but a thumbnail would obviously be better.

3. A way to extract tags from the file, so they can be added to more than just the date tag.

So… How to do this?

Option 1: I’d add a new PS_IMPORTERS_DIR. This would be a path to a directory containing files named either the file extension, or the mimetype (so psd.js or application/photoshop.js for PSD imports). This JavaScript file would have to export a set of well-defined functions to do the above 3 steps. I suspect this would be simplest to do with the PhotoStructure for Node edition, as that already requires the compiler toolchain to be set up on the box.

Here’s a sketch of the API I’m thinking you’d need to implement:

import { ExifDate, ExifDateTime } from "exiftool-vendored"

export interface PhotoStructureImporter {
  /**
   * Given a path to a file, try to pull out the "captured-at" time, or time
   * that the file was initially created.
   *
   * Ideally, a date and time, and possibly a time zone, can be returned,
   * encoded as an ExifDateTime object.
   *
   * If no time can be extracted with confidence, an ExifDate object can be
   * returned.
   *
   * If no "captured-at" time can be extracted, return `null`.
   *
   * @see https://photostructure.com/faq/captured-at
   * @see ExifDateTime
   * @throws Implementations should raise an error if the file is invalid.
   */
  extractCapturedAt(nativePath: string): Promise<ExifDateTime | ExifDate | null>

  /**
   * Given a path to a file, render as a JPEG.
   *
   * Rendering at a resolution higher than 5k (5120 × 2880) is probably
   * unnecessary.
   *
   * @throws Implementations should raise an error if they cannot render the
   * file.
   */
  renderAsJpeg(
    sourceNativePath: string,
    destinationJpegNativePath: string
  ): Promise<void>

  /**
   * Given a path to a file, extract tags suitable for your PhotoStructure
   * library, like `[["Who", "Rogers", "Fred"], ["Where", "France", "Paris"]]`
   *
   * @throws Implementations may raise an error if they cannot extract tags.
   */
  extractTags(nativePath: string): Promise<string[][]>
}

Option 2: something else y’all come up with that is easier to manage

[sandeep]

I like this approach!

[bdillahu]

Any option to just import the file with ‘generic’ values for all that and only show it in ‘folder’ view?

[mrm]

I could just take a list of file extensions, perhaps, and just show a generic file icon (with the basename?)

• I just found out that Photoshop, if in “compatibility mode,” will embed a PhotoshopThumbnail that I can extract (which is all the psd.js library is doing)

• I found that Alpine does indeed package up ghostscript, so I could add that to the docker image to add illustrator and PDF support, and ask users to install ghostscript if they wanted support.

• Alpine also has a recent graphicsmagick build, and holy moly does it support a lot of formats.

• There’s a python library that generates previews for a ton of different formats. It delegates to libreoffice to do word/excel conversions, inkscape for SVG, and poppler-utils to do PDF.

I’d basically use ghostscript, graphicsmagick, libreoffice, gimp, and inkscape as helper applications just like I currently use ffmpeg and vlc: if PhotoStructure can find them, and the version is recent(ish), I’ll automatically enable support for those file formats.

Does that sound like reasonable default behavior? I already have a copyToLibraryMimeTypes setting which defaults to “image/" and "video/”, but Photoshop’s MIME type, according to ExifTool, is application/vnd.adobe.photoshop, so I’d need to add that to the default list. Hilariously, depending on the file format version of the psd, it might also be image/vnd.adobe.photoshop, application/x-photoshop, application/photoshop, application/psd, or image/psd…

[bdillahu]

I think this sounds like a good plan - both ghostscript and graphicsmagick have good reputations, have been around for years and a ton of work has gone into their ability to handle file types - seems like leveraging that would be a good plan with less “roll your own” aspect.