ciid is a utility to derive a chronologically sortable, unique identifier for images.

Usually, digital cameras and phones assign file names to images with a sequence of only 4 digits (e.g. IMG_1234.dng). Those names will easily clash for any sufficiently large amount of images.

ciid tackles this problem by deriving a hash from the image buffer. Additionally to being able to derive an identifier that is very unlikely to clash, this hash can later be used to check the integrity of the image content.

Some image processing programs update metadata of files (e.g inline JPEG- previews, tags, modified date). The resulting ciid will be unaffected from those changes, since only the actual image buffer is hashed. This has the nice side-effect that proprietary camera RAW file formats and converted .dng files will yield the same identifier most of the time.

Here's how a resulting identifier looks like:

01234567890123-a0b1c2d3e4f5a6b7c8d9e0f1a2b3c4d5e6f7a8b9c0d1e2f3a4b5c6d7e8f9a0b1
└─────┬──────┘ └──────────────────────────────┬───────────────────────────────┘
  timestamp                         hash of image buffer

The first part of the identifier encodes the creation date of the image (a unix timestamp with millisecond precision), while the second part is a hash (SHA-256) based on the contents of the image buffer.

Following criteria were considered when choosing the identifier:

• have sensible alphabetical ordering on file systems (timestamps with a higher value should appear strictly after lower ones)
• encode information sufficiently compact
• be safe to use in URLs

Download and run the installation script:

$ curl -s https://raw.githubusercontent.com/pablosichert/ciid/master/bin/install.sh | bash

• Rust toolchain
• exiftool
• LibRaw

For help with installing the dependencies, have a look at the install script.

Install the ciid binary onto your system via cargo:

$ cargo install ciid

$ ciid [FLAGS] <file path>...

Why do we encode the timestamp as 01234567890123 instead of e.g. 2009-02-13 23:31:30.123? The timestamp represents an unambiguous¹ single point in time, whereas the date string needs to be contextualized with a time zone. That means that you would either need to annotate the date string with a time zone or change the file name every time you are on a system which uses a different time zone.

Apart from that, the former encoding is more compact.

While unfortunately it's not easy to derive the actual date from the timestamp just by looking at it, you can compare two timestamps chronologically by sorting them by value.

¹ ignoring leap-seconds.

The changelog format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

• The timestamp used in the identifier is no longer a byte sequence encoded using base32. Instead, it uses the plain decimal representation of a unix timestamp with millisecond precision. The reason for this change is that the file system UI (at least on macOS) does not sort file names strictly chronologically. E.g. files with names 01, 0a, 10, a0 should preserve the order, but get presented in order 0a, 01, 10, a0.
• The hash used in the identifier is now encoded using lowercase hex instead of base32, in accordance with how common tools encode SHA-256 hashes.

• The new CLI option --timestamp-digits can be used to control how many digits the timestamp should carry at a minimum. Missing zeros will be padded from the left.