ghotptp.txt is based on

    http://gtps.math.cmu.edu/cebrown/landau/ghotptp.txt.gz

created by Chad Brown about October 2006, with 8789459 bytes.
It has 13788 lines, one thf sentence per line.

I selected the longest line, which is line 7318 with 7015 characters, into
ghotptp-longest.txt.  Then I ran

hotptp-yl-parser-verbose < ghotptp-longest.txt > ghotptp-longest.tree

egrep -v '^ *<.*>$' ghotptp-longest.tree | awk '{i=match($0,"[^ ]");s=substr($0,1,(i-1)/2);print s $2;}' > ghotptp-longest.tokens

The second command extracts the tokens with their indentation, one token
per line, into

    ghotptp-longest.tokens.

There appear to be 1296 tokens.
Matching parens and square brackets line up under each other.
Relative indentation of unrelated tokens (e.g., ^ and @) is not easy
to interpret; however, relative indentation of various occurrences of
the same token is reliable.
Thus, it is evident that @ is being parsed as left-associative.

Other visual checking should be done to confirm that the sentence is being
parsed as intended.

--Allen Van Gelder, Nov. 3, 2006.