Intercept the exiftool interactive API

A popular and extraordinarily thorough program for reading and writing image metadata, exiftool supports an interactive mode that accepts commands through stdin. This mode is useful for batching commands without the overhead of spawning a new Perl process for each command.

The digiKam, a cross platform image management and editing application, uses exiftool as one of its metadata providers, and in the beta of version 8, also supports using exiftool for writing image metadata.

For a project that I'm working on, I wanted to log an interactive session between digiKam and exiftool. This can be accomplished without modifying either the digiKam or the exiftool source code, by introducing an additional program that sits in between, taking input on stdin, logging it, and passing it along to exiftool. As replies come back from exiftool, these too are logged, and then passed along to digiKam.

To achieve the effect, we need to introduce a bidirectional conduit. With Node.js.js, we can do this with a combination of the node:readline and node:child_process modules:

CopyRequire modules: copy code to clipboard
const readline = require('node:readline');
const { spawn } = require('node:child_process');

First, we need to capture data from stdin. This input is line oriented, and therefore node:readline (docs) makes this easy to work with. We are exclusively interested in input. For simplicity, I am using the callback API.

CopyConfigure node:readline: copy code to clipboard
const rl = readline.createInterface({
    input: process.stdin,
    output: null,
  });

Now we can spawn exiftool as a child process, and wire up its stdin, stdout, and stderr:

CopyConfigure sub process spawn: copy code to clipboard
const exiftool = spawn(exiftoolPath, args, {
    stdio: ['pipe', 'pipe', 'pipe'],
  });

The stdio key accepts an array of file descriptors, and these correspond to stdin, stdout, and stderr. We want access to all three steams, so we specify pipe for each.

Last, we want to wire up our events, so that we can log and pass data to and from the child exiftool process:

CopyLog and redirect output: copy code to clipboard
// `fs` module required, and `logPath` defined elsewhere
  const logAndWriteTo =
    (handle = 'stdout') =>
    (s) => {
      fs.appendFileSync(logPath, `${s}`);
      process[handle].write(s);
    };

  exiftool.on('spawn', () => { // <1>
    rl.on('line', (cmd) => { // <2>
      fs.appendFileSync(logPath, `${cmd}\n`);
      exiftool.stdin.write(cmd); // <3>
      exiftool.stdin.write(EOL); // <3>
    });
  });

  exiftool.stdout.on('data', (s) => logAndWriteTo('stdout')(s)); // <4>
  exiftool.stderr.on('data', (s) => logAndWriteTo('stderr')(s)); // <4>

In the above code snippet, we setup the following:

1. Setup an event handler that runs after the exiftool binary successfully runs.
2. Setup an event handler for stdin where we log incoming commands to exiftool.
3. Forward each command to the child process. We make use of the eol npm module to send a proper line ending, as readline removes this.
4. We add an event handler on each of exiftool.stdout and exiftool.stderr to capture, and forward, output from the child process.

It's important to note that, the end of line termination sequence might not be a newline. Because exiftool uses a newline, this implementation is quite straightforward. But nulls or some other sequence is possible in the wild.

The following is output from the interceptor script:

Copycopy code to clipboard
-stay_open,true,-@,-,-common_args,-charset,filename=UTF8,-charset,iptc=UTF8
-json
-G:0:1:2:4:6
-l
/path/to/a/file/20140220-103514.jpg
-echo1
{await0000000001}
-echo2
{await0000000001}
-echo4
{ready}
-execute
{await0000000001}
{await0000000001}
[{
  "SourceFile"...
...
{ready}
-stay_open
false

To use this in practice, it is crucial that digiKam can actually execute the script, which is prefaced with #!/usr/bin/env node and therefore node must be in one of the paths in the PATH environment variable.

By default, digiKam searches your PATH environment variable find the exiftool binary. However, your login shell PATH variable might not be the same as the PATH variable that digiKam starts with. For example, on OS X, the environment of the launchd process is the one that digiKam starts with.

On OS X, you can start digiKam with your shell's environment by running open /Applications/digiKam.org/digikam.app.

The complete code is available from the exiftool-intercept repo on GitHub.