Oh ... so that is what a Pratt parser is. Thanks for the nice clear writeup. But methinks it should be called a Wirth parser. Nicholas Wirth invented the sam

Oh ... so that is what a Pratt parser is. Thanks for the nice clear writeup. But methinks it should be called a Wirth parser. Nicholas Wirth invented the same thing in the 1970's. He initially used it to parse his first language, Pascal. He drew them as Syntax diagrams: https://en.wikipedia.org/wiki/Syntax_diagram https://en.wikipedia.org/wiki/Syntax_diagram The loops that characterise Pratt are evident in those diagrams. You can read his book on Pascal here: https://www.research-collection.ethz.ch/bitstream/handle/20.500.11850/68910/eth-3059-01.pdf https://www.research-collection.ethz.ch/bitstream/handle/20.... At the end you will find the syntax diagrams for the entire language. The compiler was derived from those diagrams in the straight forward way described in the article. And everything the article says is true, if you are going to manually write a parser, they are by far the easiest starting point. And I don't doubt error messages and error recovery are much easier too. Parser generators don't give you the same flexibility to play with a parsers guts. But you do lose some things. A parser generator checks for ambiguities, something you can easily miss in a hand written parser. A parser generator also enforces a clean split between the parser and whatever you are doing with the syntax tree, whereas hand written parsers lead you toward a ball of mud where one lump of code has all the tasks intermixed. And there are LR parsers out there that support the loops and priorities that raw BNF productions make so hard. Like mine :D http://lrparsing.sourceforge.net/doc/html/ http://lrparsing.sourceforge.net/doc/html/