progressive rendering: use the previous word count for the word rate

[aeschli]

For #198558

I assume the goal of progressive rendering is to spread out rendering of words so there are no pauses

When the word rate is computed, we currently use the new content to compute the rate. But using that rate we would have already printed the full content, which just arrived.

IMO we should use the previous content.

[roblourens]

I don't follow- this is computing the rate that text has loaded, so all the words over all the time. You are excluding the latest words, but still dividing over the full range of time. Why is that better?

[roblourens]

But using that rate we would have already printed the full content

The goal is just to estimate the average rate at which words are loading, and render the next block at that rate.

[aeschli]

Let's say we receive 10 new words every second. At second 1 we receive the first 10 words. At second 5 we had 40 words and receive another 10 words.
Before, the average was computed on the new total: 50 words in 5 seconds = (10w/s). Shorty after 5 seconds, the evaluation would be that we should have printed 50 words by now, so we would print the new 10 words, in one go, to reach the 50 words.
But really we only had 40 words to render at second 5. So by setting the rate to 40 words in 5 seconds (8w/s) we basically slow things down to evenly stretch out the printing.

[roblourens]

I see- I'm not worried about rendering at a nice even rate. I'm trying to compute the rate that words are coming in, and try to match that rate so that we don't fall too far behind. I think we should always start printing faster, if we can.