Re: [Nano-devel] nano bug #54928: a syntax file should be fully read only when the syntax gets actually used

[Brand Huntsman]

On Wed, 14 Nov 2018 00:14:17 -0700
Vlad Vladov <address@hidden> wrote:

> Is anyone working on this right now? If not I would be interested in
> taking a stab at it. Any suggestions where in the code I should look
> (I am new to this codebase)?

parse_rcfile() in rcfile.c with syntax_only==TRUE. The syntaxtype struct used 
by live_syntax is in nano.h, you will need to copy and store (at the end of 
parse_one_include()) the filename for the syntax in syntaxtype and once loaded, 
free it and set field to null. A single conditional between "magic" and 
"comment" can abort if (syntax_only && only_parse_syntax_headers).

Opening a file in a new buffer or renaming a file will need to load the syntax 
if not already loaded. Look for code that sets the syntax field in an 
openfilestruct. Syntaxes are probably currently read before any files from the 
command line are known, so that will be fun figuring out which initial syntaxes 
to load. Maybe just initialize an only_parse_syntax_headers global to false 
which causes parse_rcfile() to only read headers during load, and then flip it 
to true when finished reading in all syntax headers, at the end of 
do_rcfiles(). The command line files are probably loaded with the same code 
used to open a file in a new buffer, so they might just work without any extra 
code.

The actual loading can be done with parse_one_include(), but you need to ignore 
"syntax", "header" and "magic" commands since they already executed, and 
calling them again will cause problems. The live_syntax global will need to be 
set to the syntax you are loading before calling parse_one_include().

You could add a reference count to syntaxtype and increment everytime a buffer 
uses the syntax and decrement when buffer is closed. When count reaches zero 
you could free the regexes. But that adds more complexity for a case that will 
rarely trigger. So I'd just leave the syntax loaded even if no files are still 
using it.

Syntaxes can also be declared inside nanorc, these will need to be fully loaded 
at startup. I believe syntax_only will be false while loading these, so you 
might not have to do anything special, other than verifying they still work.

Thank you for wanting to implement this, it will help a lot in reducing memory 
usage.