• AndyC (11/11) Feb 04 2015 The vision says "We aim to make the standard library usable in
• Andrei Alexandrescu (3/4) Feb 04 2015 It will allocate, but without creating garbage. Reference counted
• Jonathan Marler (31/42) Feb 04 2015 It depends on the implementation, but it should be possible to

"AndyC" <andy squeakycode.net> writes:

The vision says "We aim to make the standard library usable in 
its entirety without a garbage collector. Safe code should not 
require the presence of a garbage collector"

I was just playing with std.json, and this part got me to 
thinking.  How will we parse a string into a json struct w/out 
allocating memory?

And if I allocate memory w/out the GC, then is the caller 
responsible to free it?

Things like string format and file read are gonna have to alloc 
memory, yes?  How will we make std.* usable w/out the GC?

-Andy

Andrei Alexandrescu <SeeWebsiteForEmail erdani.org> writes:

On 2/4/15 1:51 PM, AndyC wrote:
 How will we parse a string into a json struct w/out allocating memory?

It will allocate, but without creating garbage. Reference counted 
strings are the answer. -- Andrei

"Jonathan Marler" <johnnymarler gmail.com> writes:

On Wednesday, 4 February 2015 at 21:51:55 UTC, AndyC wrote:
 The vision says "We aim to make the standard library usable in 
 its entirety without a garbage collector. Safe code should not 
 require the presence of a garbage collector"

 I was just playing with std.json, and this part got me to 
 thinking.  How will we parse a string into a json struct w/out 
 allocating memory?

 And if I allocate memory w/out the GC, then is the caller 
 responsible to free it?

 Things like string format and file read are gonna have to alloc 
 memory, yes?  How will we make std.* usable w/out the GC?

 -Andy

It depends on the implementation, but it should be possible to 
parse an entire JSON file without allocating memory.  Instead of 
creating new memory for every string in the JSON, you can just 
save a slice to the JSON itself. However, this will not work if 
the user intends on freeing the JSON memory later.  Either way, 
this is an option I believe the library should provide.  In some 
cases it's more efficient to read the entire JSON file into 
memory and just save slices into that memory.

Note: If you need to escape strings you will have to modify the 
original JSON in memory, which may or may not be ok.  Again, the 
library should provide a way to specify this.

You could also allow the user to pass an allocator to the JSON 
library.  This allows memory management to be implemented by the 
user instead of forcing everyone to use whatever memory 
management the JSON library uses.

One last tip.  If you don't care about preserving the original 
JSON, you can re-use the JSON memory by "moving" all the strings 
you need for your in-memory JSON at the begining of the JSON 
buffer.

Original JSON memory:
{
    "myid" : "hello"
}

"in-memory" JSON

myidhello
^   ^

All the strings are moved to the beginning of the JSON buffer.  
Then you can decide if it's worth it to free the original JSON 
buffer and copy what you need to a new buffer or just keep the 
original buffer and ignore the wasted space left over.