Using (save) is very, very slow for some reason. (load) on the other hand, is quite fast.
I understand (save) is doing slightly more work in the following example, but the speed difference is staggering. And, for large lists... the speed difference makes me want to use (write-file) and do conversions, because (save) is unbearable.
Take the following example:
(set 'seqx (sequence 1 1000000))
(println (date) "-Pushing sequence to list")
(dolist (x seqx)
(push x foo -1)
)
(println (date) "-Saving list as foo.lst")
(save "foo.lst" 'foo)
(println (date) "-Writing the list as foo.txt")
(write-file "foo.txt" (string foo) )
(println (date) "-Finished")
When run, it produces this output:
Quote
Sun Apr 25 15:07:15 2010-Pushing sequence to list
Sun Apr 25 15:07:16 2010-Saving list as foo.lst
Sun Apr 25 15:07:43 2010-Writing the list as foo.txt
Sun Apr 25 15:07:46 2010-Finished
Pushing the sequence to a list - 1 seconds
Saving the list (save) - 27 seconds
Writing the list (write-file) - 3 seconds
And, this gets exponentially longer the greater the size and depth of the list I am saving.
The function 'string' writes to memory, while 'save' opens a file then does file I/O piece by piece.
I have now changed 'save' to convert everything to memory first, then write to a file in one swoop. This makes 'save' now actually slightly faster than (write <file> (string <symbol>)) :-)