Filling assembly machines

[Factorio2016]

translator
Why the manipulator puts into a build machine at first for a double portion of the first required resource, then a double portion of the second. Why? First one portion of the first resource, then one portion of the second resource.
So save time filling Assembly machines. And increases the rate of production at the early stages of the game.

[Hexicube]

Better translation:

Why do inserters fill two request's worth of a single item before moving on to the next item required? It would make more sense to do a single request's worth of all items before moving on to the second.

[Koub]

Legitimate question. I alse asked myself the same thing, but after 10 minuets, when you have 10s of assembling machines producing, you just forget it.

[Factorio2016]

To forget? why?
It's a simple fix, well I think so.
10 minutes maximum to correct the code.

[ssilk]

Well sounds legit on the first view and makes much sense.

But the devil is in detail: If you have two inserters inserting in an assembly: How should they work then? Under which circumstances should an inserter insert what? Consider, that the second inserter already could have something in the hands and is currently moving to insert it.

I think that is one of the reasons, why it is, as it is, cause it was the easiest (and fastest) way to calculate that problem.

Of course that doesn't mean that if you have a better algorithm for all the cases inserters can insert doesn't make sense.

[Koub]

I don't know how the request for the actual needed goods is sent from assembling machine to inserters, but I have an algorythm that can mostly "fix" that strange behaviour :
Give a weight for each component needed to produce the current recipe depending on how many have already been inserted.
For example a recipe that needs 5A, 8B and 1C :

- Assembling machine has 0A, 0B and 0C requests 100% A, B, C
- Inserter grabs a B
- When B is inserted, assembling machine asks for 100% A and C and 87.5%B.
- If given the choice, it will take either A or C, else B. (from most needed to least needed) let's assume there's A in front of the inserter.
- A is taken, and inserted. The needs are now 80%A, 87.5%B and 100%C.
- ...

And so on. As soon as the needs are 0% for all components, the machine can start crafting. It will have taken 5+8+1 = 14 insertion, +eventually 3 per additional inserter (if unlucky)
With the current algorithm, there would have been 5x2+8x2+1= 27 insertions before the craft starts.

- ...

If inserters are faster than the machine crafts, the machine stops asking for additional components when all needs are at -100%

This has durable effect only if the assembling machine is faster to craft than inserters are to insert. Else, the only effect is at the start of the crafting, until inserters have inserted enough to start the next craft before the assembling machine has finished the current one.

[ssilk]

Someone could argument: An inserter should try to insert the same item type as long as possible, cause that way his belts are better occupied (even the last machine in a row has a chance to grab items).
Others could argument: Why change? On the long run it doesn't matter in which order the items are inserted, it's only for the first produced item, where that can be optimized.