Clojure is designed with enough extensibility to modify the printer and reader to cover this case.
You can define a custom print strategy for these types by extending the print-dup multimethod. If you print it as a tagged literal, you can also define a custom reader that can read it back as the appropriate type. Or you can just happen to write it as a form that constructs a Java object if that's possible. (I don't know the Joda classes well.) Print methods typically look like this: (defmethod print-dup org.joda.time.DateTime [date-time ^java.io.Writer w] (.write w (format "#org.joda.time.DateTime[%d]"))) That particular format should produce a string like #org.joda.time.DateTime[1462809085214] which when read will call the DateTime constructor with the long value. Or you could write your own custom tagged literal format and define and install a reader for that tag that refers to a function that does whatever you need. And I would recommend never using Java serialization in either Clojure or Java. :) On Monday, May 9, 2016 at 9:44:36 AM UTC-5, Simon Brooke wrote: > > I'm working on some complex analysis where it takes about twelve hours to > construct the model to be analysed. Once it's constructed I'm doing various > interactive things to explore it, but I can't currently persist it in a > form in which it can be read back in, so that when the REPL session ends it > is lost and I need to run the construction process again. > > When I was creating the system I had thought, airily, 'well, Clojure is > Lisp so all I need to do is spit the structure out to a file and slurp it > back in again'. Unfortunately, of course, Clojure is only mostly Lisp. > > My problem is clj-time objects, both date-time and interval. These are > essentially Joda time objects, and when they are printed, what you get is a > string representation of an object reference. When the reader reads it > back, what you have is a string: > > user=> (def foo (tc/now)) > > #'user/foo > > user=> (spit "/tmp/saved-time" (with-out-str (pr foo))) > > nil > > user=> (def bar (slurp "/tmp/saved-time")) > > #'user/bar > > user=> (= foo bar) > > false > > user=> (type foo) > > org.joda.time.DateTime > > user=> (type bar) > > java.lang.String > > > So, my question: in this particular case, what is the simplest means of > writing my data structures (which, apart from the clj-time objects, > comprises deeply nested maps of keywords, strings and numbers - all things > which the Clojure reader and writer can cope with) to disk in a form in > which they can subsequently be read back? > > > There are two obvious solutions: one is to have a function which walks the > data structure and returns an equivalent data structure in which each > clj-time object is replaced by a function invocation which when invoked > creates an equivalent clj-time object; the other is to use Java > serialisation to serialise the whole structure as in the answer to this > stackoverflow question > <http://stackoverflow.com/questions/7701634/efficient-binary-serialization-for-clojure-java> > . > > > Which of these would people go for? Or is there some other solution I have > not thought of? > -- 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 --- You received this message because you are subscribed to the Google Groups "Clojure" group. To unsubscribe from this group and stop receiving emails from it, send an email to clojure+unsubscr...@googlegroups.com. For more options, visit https://groups.google.com/d/optout.
