Real World Haskell was the first book that made it to a wider audience. And while was is full of useful information, it is a very uneven book that works badly as an introduction to Haskell. We once had a reading group where we read RWH, most colleagues dropped out after the first 3-4 chapters.
In comparison, LYAH was a book that was far more straightforward as an introduction. And I know a few people that have gone through it cover to cover.
Although it's been a while since I read it, I really liked Graham Hutton's 'Programming in Haskell'. It was short & clear.
This is a bad statistic to use on its own. Imagine if I said "Majority of people I anecdotally hear gave up on web development seem to mention HTML". See the problem?
LYAH is not intrinsic to Haskell programming.
Imagine a hypothetical universe where all Haskell books are equally good, but some are more popular than others (maybe they have well known authors, or catchy titles, or are available for free online, or something). In particular, imagine Haskell Book Number 7 (HB#7 for short) gets 99% "market share".
Then if someone says "most people who gave up on Haskell used this book", you should not conclude that it reflects badly on the book. It's just a consequence of the book being popular, so it's the one people tend to try to use, so it's "the one that didn't work for them".
If you want to learn whether the book is better or worse than average, you need to split people into two groups: one group used this book, the other group used some other book, and compare which group does better. (Ideally, you should assign people to a group based on a coin flip, but even just looking at what happens "naturally" without your interference is better than nothing.)
Back to our non-hypothetical universe: LYAH does not have 99% market share, but it's pretty popular. I heard of it, and I've never looked for a book on Haskell. So to some extent the same logic applies.
As you say, though, LYAH is almost certainly not at 99% market share. It's definitely not at the 100% market share HTML has in Web development.
The universe is too complex to design a statistically-valid study for everything you want to know (though I definitely understand the urge to do that). At the end of the day, anecdotal evidence may not be great, but it's often better than nothing, and nothing is often the only other practical option.
True. The trick is to sort of mentally keep track of how much better than nothing it is, so you know which pieces of anecdotal evidence you can trust with only 40% confidence, and which you can trust with something more like 90%.
But yeah, definitely WAY better than nothing.
I also recommend reading these web documents as a supplement:
http://dev.stephendiehl.com/hask/
https://wiki.haskell.org/All_About_Monads
https://wiki.haskell.org/Typeclassopedia
https://www.fpcomplete.com/blog/2017/09/all-about-strictness
I am now reading Haskell High Performance Programming (https://www.packtpub.com/application-development/haskell-hig...), and while the book is more intermediate (it assumes that you can find the documentation and read the Haskell code yourself), I think it is a good overview of how to use modern Haskell in real projects.
Also this is one of the best references while you're learning (not good as a tutorial, but chock full of invaluable resources): http://dev.stephendiehl.com/hask/ . TBH I wish every language had an equivalent of this guide.
The beginner material in the Haskell WikiBook is pretty solid: https://en.wikibooks.org/wiki/Haskell
Fairly sure it was recommended in here in the past and that's how I came to know it, can't give credit because I sadly don't remember.
[1]https://www.seas.upenn.edu/~cis194/spring13/lectures.html
While not purely functional, I also recommend JacaScript Allongé.
Both are free online!
It's not. It's a terrible starting resource. Majority of people I anecdotally hear gave up on Haskell seem to mention LYAH. Yes the material is there, but it's not presented in a way that generally sticks and fits pieces together. I guarantee someone will reply to this thread and say it worked for them, but in general people seem to be unsuccessful with it.
There are better options out there.