use v8.serialize(), v8.deserialize() for snapshots?

[isaacs]

This might be a terrible idea, since part of the value in tap's snapshots is that they are mostly human readable (provided you aren't snapshotting objects containing strings with \ in them, since those get quadruple escaped, which is brutal when a string contains \\ and gets snapshotted as {"key": "\\\\\\\\" }).

But maybe we could speed things up a lot by taking advantage of the fact that snapshots usually match.

In --snapshot mode, t.matchSnapshot(obj) would write v8.serialize(obj).

When comparing, we could check that v8.serialize(obj) === snapshot[key], and if they don't, then diff format(obj) vs format(v8.deserialize(snapshot[key])).

That'd be much smaller strings to compare in many cases, and snapshots files that don't grow so massive.

Otoh, it makes diffs useless. So, maybe terrible idea.

[isaacs]

Eh, it'd only cut the strings about in half, and we'd lose post hoc analysis. Bad idea.