See also @BillJ's answer, which is almost identical to mine, except that he doesn't cite CGEL and phrases it in his own words.

CGEL does not cover this usage specifically where you have a full sentence or quasi-sentence used as an adjective. And this is not reported speech. I have provided numerous examples of this.

Could you explain to me why you think 'I don't know how many' is an adjective? What makes your analysis correct, and CGEL's wrong? (Apologies for the bluntness.) It may well be that CGEL's analysis is wrong, or at least not a definitive source on this matter.

It is not that CGEL's analysis is "wrong". It is a sentence premodifier. CGEL just don't analyze it thoroughly or give other typical examples. As used in the OP's utterance, it is adjectival. If you move it to the end of the sentence, it then becomes something else as in your i). The occurrence of three-hyphen phrases is well described "in the literature"; Longer phrases or sentences are not. However, it is very, very common in speech and also used in writing. See my answer and the link to Sven Yargs on a similar question.

Thanks for the bluntness. No beating around the bush. :)

@Lambie, after searching the Internet for some time, I chanced upon several papers that deal with these types of constructions, which are called 'syntactic amalgams'. To my knowledge, these constructions were first analysed in detail by George Lakoff (a former professor of linguistics at UC Berkeley) in 1974; he defined 'syntactic amalgams' as 'sentences which have within them chunks of lexical material that do not correspond to anything in the logical structure of the sentence; rather they must be copied in other derivations under specifiable semantic and pragmatic conditions'. (Thread, 1/?)

Hey, Ben, That's GREAT. And a good name for them. syntactic amalgams. Wouldn't want those in my teeth. My only disagreement is that I still contend they function adjectivally in the cases I describe and in the OP's sentence. You know, you'd think prof, editors would have stuff about this on the internet.

The syntax of such constructions is still a matter of considerable controversy. On the one hand, some sources (such as CGEL) contend that 'I don't know' in the sentence 'He committed I don't know how many crimes' is an irregular modifier of 'how many crimes', and that the syntactic function of 'I don't know how many crimes' is that of direct object; while others argue that 'I don't know many' is a complex determiner of 'crimes' (M. Guimaraes), with 'crimes' being the latter. I have yet to see any sources that argue that 'I don't know how many' has the role of an adjective, but... (2/3)

Ben, I believe in following the logic of the language. CGL does not really even deal with these to any extent, especially where they are hyphenated. Why do you and others here act AS IF you don't recognize the kind of examples I have in my answer?? I don't understand that.

And now you repeat there is controversy about the "construction". The entire issue disappears when you add hyphens and make them premodifiers of the nouns.

I disagree that there is controversy. For example: "This is an example of one of those damned-if-you do-damned-if-you-don't situations." Who would argue about that? The OP's question is not hyphenated because wherever they fished it out of, didn't. That does not mean that most of these are not hyphenated because they are.

The point is not how most native speakers would say it. The point is that that utterance was said that way and it follows a pattern of hyphenated sentence premodifiers.

In .

@Ben "The party will be at you know where" is not in the pattern we are discussing. This is: "John invited a lot of people to a you-can-guess what-kind-of party."

There are clear premodifiers, and others that are not. He is clearly a spend-now-cry-later guy.

I think it would be best for us to end this conversation here, since it seems that neither of us will ever be convinced by the argument of the other. Good day (or evening). (Again, apologies for the curtness.)