>> So, what are people's thoughts? > > Trying to use them in Common Lisp has frustrated the crap out of me. > The only library I know of that promulgates them seriously is CL-SQL > and the gymnastics you have to do to install the reader macros are > frustrating. Another library I tried to use I couldn't get to work > at all (units, I think).
AllegroGraph uses ! as a reader macro to introduce URIs and RDF literals, which means they appear quite frequently in user code. You can write !rdf:type !<http://example.com/foo> !"chien"@fr and the appropriate object (a so-called future-part) is integrated directly into the read code, just like a string or a list. This has three advantages over something like (resource "rdf" "type") — * This process happens at read-time, which means that you can use this syntax in *unevaluated* positions, such as in macros, query forms that are only read, not evaluated, etc. That can be really convenient. * The syntax is much closer to the conventional syntax used in serialization formats (simply add a '!' to the front of the usual syntax). Users who are more familiar with Java or Python don't seem to mind, which is not true when you add parentheses to your syntax. * You can raise a syntax error at read time, rather than at compile- or run-time. That's a nice feature in a dynamic language, where you might not know your code is broken until much later. It's nice for DSLs to have distinct syntax and semantics, and requiring the use of macros or functions to implicitly construct syntactic elements blurs that line. I've observed this in the discrepancy between the 'primary' Clojure types, such as set, map, et al, which have syntax, and those which do not: sorted-set or struct-map, for example. However, for all these benefits I'm still not sure whether it would be worthwhile allowing user-generated reader macros. They are absolute hell when integrating conflicting libraries — it only takes one major library to use a reader macro character, and it's pretty much verboten for all other libraries, just in case you want to combine them. Franz's Allegro Common Lisp offers a workaround for this problem — named readtables, which makes using readtables in CL much tidier. If a library's readtable does not take effect in user code until demanded, then conflicts do not occur. Code that uses libraries with conflicting syntax can be partitioned to use different readtables. If it were possible to signal to Clojure's reader which readtable to use when reading a file/stream/string, and readtables were named in a similar manner to namespaces, and could be composed, then I think user- defined reader macros would be safe and convenient. Just as in-ns and ns adjust the current namespace, they could adjust the readtable: (ns com.example.foo (:refer-clojure) (:use com.example.library1) (:readtables clojure.core com.example.library1)) I'm sure this would complicate Clojure's reader significantly. -R --~--~---------~--~----~------------~-------~--~----~ You received this message because you are subscribed to the Google Groups "Clojure" group. To post to this group, send email to clojure@googlegroups.com Note that posts from new members are moderated - please be patient with your first post. To unsubscribe from this group, send email to clojure+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/clojure?hl=en -~----------~----~----~----~------~----~------~--~---
