What does lazy evaluation mean in the context of the dplyr r package, and when should I not use it?

Question

I recently read the non-standard pricing chapter from the dplyr vignette after I came across dplyr functions that end in _ . Hearing the introduction to Haskell a long time ago, I roughly remember “lazy rating” as a term for “code rating, but only when or when it is really necessary” - and reading the lazy form of rating evaluation Hadley Wickham Advanced R seemed to confirm this memory.

What I don't seem to be able to hug is that NSE / SE (lazy / impatient evaluation) is used for dplyr and when, for example, functions such as summarize_() are used compared to summarize() .

I might have missed part of the documentation somewhere. I would be happy to narrow down the topic of my question or even step back if it becomes clear that the question is too broad or to which an answer has already been given.