Comment by rhythmvs | Hacker News Reader

Comment by rhythmvs | Hacker News Reader

Parent Full thread

A file naming convention (lightweight markup) that would allow us to store structured (meta)data right inside file names. Obviously inspired by Markdown and CSV.

We could then build lean, database-less asset management applications, while the user data (i.e. the files and their metadata) would always be portable, across platforms.

Take for example:

  J.M.W. Turner | Rain, Steam and Speed | ···· 1844.jpg
  W. Blake ···· | Newton ·············· | 1795–1805.jpg

as compared to the clutter we now must deal with:

  _IMG00123.JPG
  Turner_-_Rain%2C_Steam_and_Speed_-_National_Gallery_file.jpg

My practical use case: take snapshots of my incoming receipts, bills, etc., name the jpgs using the proposed file naming convention (including fields for VAT, net amount, etc.), put them in Dropbox, build a parser and accompanying GUI to edit file names (and their corresponding metadata; have total amounts etc. being calculated in real time), drop a link to that (web app) interface to my accountant.

It’s just an idea for HN Idea Sunday; I did a somewhat more detailed write-up:

https://gist.github.com/rhythmus/11118629

I like this. I've often had a similar issue in computational physics when handling lots of different calculation input/output files with variations in parameters. I've tended to automatically generate directory hierarchies, e.g.

rc=1.0/ecut=60/kp=4,4,4/

rc=1.0/ecut=80/kp=4,4,4/

rc=2.0/ecut=60/kp=4,4,4/

rc=2.0/ecut=80/kp=4,4,4/

but some standard way of expressing it in the file name would be great, and nicer to parse out later. There are filesystems that allow metadata, but I've never seen one really being used for that purpose. You could make some associated command line tools that are equivalent to 'ls' that allow splitting/slicing the files down different parameters.

Take a look at hdf5.

Thanks, I'll take a look, but it seems like I'd have to change the format of the files, right? Unfortunately that's not really feasible, since their format is determined by the scientific codes I use -- that's why I was attracted to the idea of using the file name. I can, however, have a go at HDF5 for code that I write myself.

You'd be interested in the Be File System (from BeOS / Haiku) for inspiration:

http://arstechnica.com/information-technology/2010/06/the-be...

It has support for database-like extended file attributes, which can be custom attributes!

For example, a `person` file would have all of the normal fields you'd expect from an address book as file attributes: http://www.birdhouse.org/beos/byte/24-scripting_the_bfs/ https://www.haiku-os.org/docs/userguide/en/attributes.html

Technical: http://www.letterp.com/~dbg/practical-file-system-design.pdf

Though you seem to be against specific file systems due to non-portability and that is understandable.

I love this idea. I rely too much on custom file naming conventions, and I always end up ditching applications for managing file metadata (for example Calibre for ebooks or several others for photos and music) because using `find` covers most of the use cases and organizing the metadata takes much time.

MetaData is already built-in to most modern file systems. On OSX (HFS Plus) you have the xattr command, and there are things for ext2, ext3, XFS, ReiserFS...

Sure, but it’s neither portable across platforms, nor user friendly (for non-techies).

Imagine managing a collection of photos (or a mixed collection of multiple file formats, like scans of receipts, and pdfs with invoices) in something like Dropbox: no access to hidden low-level metadata blocks, etc., let alone easy editing. Not to speak of cross-platform support.

File names are always visible and editable, and they’re easily parsable.

I discussed such objections here: https://gist.github.com/rhythmus/11118629#problem-metadata-p...

+1 to this idea. Organizing photos, I would like to impart date taken, location, and context (e.g., 2014-04-29; San Francisco, CA (or geolocation); "California trip"). And I'm sure other people would love other types of metadata associated with photos. And that's just image files... plenty of other metadata-needy situations (e.g., the receipts you mentioned; I have a similar problem too).

You can just put this in the EXIF data block of the image. File explorers like Windows Explorer show this meta data just like any other meta data about the file like file size.

Reply on news.ycombinator.com