eval is evil (and undocumented)

[benjie]

Ref: https://discord.com/channels/489127045289476126/498852330754801666/1205068909787746304

So apparently the eval methods are completely undocumented... Good to know.

Yes; all the eval methods have a cost, but the more specific you can be the better. .evalHas(key) causes the plan to branch such that .evalHas('foo') will have two versions of the plan - one where foo is present, and one where foo is not present. (Compare to .eval() where each individual different object will result in a new plan.)

$__inputStep.evalHas(key): two plans
$__inputStep.eval(): almost infinite plans
$__inputStep.evalLength(): one plan for each length of the list
$__inputStep.evalIs(val): two plans (one where $__inputStep's value === val; and one where it doesn't)

[benjie]

@jemgillam I think details of this was in one of the sponsors only talks; perhaps you could extract it and add to the docs?

[benjie]

Actually, we're looking to maybe remove eval if we can... So let's put this on ice.

[jemgillam]

how does tamedeval relate to eval() ?
A better question is, can one use tamedeval in place of eval() in Version 5?

tamedeval doesn't solve the problem of multiple plans being generated.

[benjie]

eval() in JS evaluates a string as code.

tamedevil makes it easier to work with eval() in a safe way.

In Grafast, input steps have an .eval() (or .evalIs(), .evalHas(), etc) method. This is unrelated to JS's eval() and thus to tamedevil; it only relates to evaluating the real value of this step during planning (and means that the plan is only valid for future requests where the evaluated value or expression matches).

[jemgillam]

@jemgillam I think details of this was in one of the sponsors only talks; perhaps you could extract it and add to the docs?

Hmmm I think I don't have access to those videos, I guess it was a Discord event?

Is discussed in June '24 working group: https://youtu.be/dqfyhM0sjoE?si=d5-t5xvqC75PZ_f_&t=1483 but that's not what you're referring to above

[benjie]

@jemgillam https://docs.google.com/presentation/d/1tcSSE1SyAchwGpPoAPDL7DbiHk1gLvODAdqL411KR08/edit#slide=id.g12471d94864_1_352 slide 43, ignore first two bullets.

[jemgillam]

Closed with #2213