Editing a todo changes its ID about todoman HOT 16 CLOSED

I still have mixed feelings about having persistent IDs. I see one disadvantage (that in my case most IDs are guaranteed to have more than one digit), but no advantages.

from todoman.

Advantages are that concurrent operations won't change the IDs, which, IMHO, adds a bit of safety:

• In terminal A run todo, walk away.
• Open terminal B, run todo, close it.
• Some time later, notice terminal A still there, and do something like todo done 3 based on what you see on-screen.

It might be a bit uncommon, but I kinda like the determinism this adds.

from todoman.

That is fair enough.

from todoman.

I'm still willing to discuss it, but I honestly, I generally favour determinism.

from todoman.

An easy fix that wouldn't change the nice cache validation flow we have (first expire, then reinsert) would be to directly derive the ID from the UID.

from todoman.

I don't think we can derive collision-resistant IDs from UIs due to the amount of digits. Consider that we'd need a two/three digit output for every UID that does not collide (think of it as a hash function).

from todoman.

We could throw in lowercase letters into the mix since those are easy to type as well.

Other options would require updating the cache in-place during validation which is somewhat complicated.

from todoman.

With letters + numbers (35 symbols), and 80 todos in the db, we'd have a 92% change of collision with two digits, and 7% with three digits. With both lowercase and uppercase letters, this goes down to 1% and 0.5% respectively.

These number are a bit high, because a single collision would make todo fail.
I'm also not convinced that the case-insensitivity of the ids will be obvious.

To get a better understaning though, what causes the dislike for persistent ids? Maybe there's an alternate solution to it it.

from todoman.

I dislike persistent IDs precisely because they fill up the space of possible IDs so much. I usually have a few open tasks where most tasks usually have a single-digit ID. With the new behavior, the tasks that I don't care about (done tasks) and will never address by ID again are assigned a low-digit ID.

from todoman.

Ah, I understand, that's really a really valid concern. Let me think if there's another solution. I take it that todo flush is not something you're fond of.

from todoman.

I consider todo flush to be a performance workaround but it works.

from todoman.

How would you feel about an autoflush setting to auto-delete done tasks? Or do you prefer to keep them around?

from todoman.

I personally have no use for them after I'm done, but we shouldn't assume that the avg user feels about that the same way.

from todoman.

Right, I meant, autoflush with default=False.

If this sounds okay to you, I'll prepare a PR, but if you actually want to keep them around, we can go back to brainstorming.

from todoman.

Since flush also wipes the cache I imagine that this slows down everything a lot.

from todoman.

It does. 😞

Back to the drawing board it is!

from todoman.