Video file names in standard format

[becky-gilbert]

This PR updates the video file names in the record package so that they match the format used by EFP and expected by the lookit-api and and the AWS S3-Lambda trigger.

• Consent videos: consent-videoStream_{study}_{frame_id}_{response}_{timestamp}_{random_digits}.webm
• Non-consent videos: videoStream_{study}_{frame_id}_{response}_{timestamp}_{random_digits}.webm

This naming system requires two arguments from the plugin that calls recorder.start:

1. whether or not this is a consent recording, which is used for the prefix
2. the "trial_type" (plugin name), which is used to create the frame_id value

The frame_id value is the same one that lookit-jspsych uses in the response sequence field, and that's used in the lookit-api to identify trials. The format is: {trial_index}-{trial_type}. Trial types are not necessarily unique across the experiment, but the trial indices are.

Implementation note

Our previous file naming system used a set of pre-defined strings as an argument in the recorder.start method: "consent" | "session_video" | "trial_video". It would be nice to still have a set of pre-defined strings for the new trial_type string passed to recorder.start. But I don't think that's makes sense now, since our trial recording extension is meant to be used with any jsPsych plugin, so we'd have to maintain a long list of all (currently ~40) jsPsych trial types ("html-button-response", etc.). So this argument is just an unconstrained string. I'm open to suggestions for ways of constraining or validating it.

TO DO

As discussed here #36, we would like to give researchers the ability to assign their own unique/meaningful trial IDs, and use these index-type IDs as a fallback. But in the interest of meeting our first MVP, I'm just implementing the index-type solution for now, and we can add the support for researcher-provided trial IDs later.

[changeset-bot]

🦋 Changeset detected

Latest commit: f6b667b

The changes in this PR will be included in the next version bump.

This PR includes changesets to release 1 package

Name	Type
@lookit/record	Patch

Not sure what this means? Click here to learn what changesets are.

Click here if you're a maintainer who wants to add another changeset to this PR

[okaycj]

This looks good.

[okaycj]

Does the type need to be converted to a string?

[becky-gilbert]

Argh, you're right, this is a class name, not the 'trial_type' string 🤦‍♀️ Fixing it now.

[becky-gilbert]

You can get the current plugin name like this: this.jsPsych.getCurrentTrial().type.info.name, but for some reason the jsPsych types produce errors in that line (info property is missing from the jsPsych plugin class interface). So I created a new interface that extends the jsPsych plugin interface.

I doubled checked in the jsPsych code that plugin classes must be created with PluginInfo, and PluginInfo must have a name property (string). Also, when running a trial, the jsPsych code assumes that a trial's pluginInfo value contains a name. So if a plugin class doesn't have a value for info.name, then I don't think it would work.

Thus I don't think we need to do any error catching here for the existence of info.name on the class, but I'm open to suggestions if you disagree. If so, maybe we could put in a default value for the plugin name, or throw an error?

[okaycj]

Would the fact that TrialRecordExtension is not a plugin change anything?

[becky-gilbert]

Nope, the fact that this is an extension is the reason why getting the plugin type here is tricky. The TrialRecordExtension can run over any plugin, and the plugin info isn't automatically accessible from within the extension code. But since we're waiting for the plugin to load before we start recording, we can also ask jsPsych at that point what the current trial type is (this.jsPsych.getCurrentTrial().type). This gives us the name of the plugin class (e.g. HtmlKeyboardResponsePlugin), and from that we can use .info.name to get the plugin/trial type string (e.g. "html-keyboard-response").