ls-sfparser is a Structured Fields (RFC 8941) parser written in C.
Features:
- Simple streaming API
- Simple to include in your projects
- Low on resources
- Portable
- Permissive MIT License
The parser is invoked with a single function, ls_sf_parse()
. The function is passed the top-level type (Item, List, or Dictionary), the input string and
its size, and the callback function:
/* Parse HTTP Priority Parameters: */
const char *input = "u=3, i=?0";
int ret = ls_sf_parse(LS_SF_TLT_DICTIONARY, input, strlen(input), my_callback,
&prio, NULL, 0);
if (ret == 0)
printf("parsing successful\n");
The callback may look like this:
/* Return 0 on success, any other value to stop parsing and return with failure. */
static int
my_callback (void *user_data, enum ls_sf_dt type, char *str, size_t len, int off)
{
struct prio *const prio = user_data;
/* --- 8< --- snip --- 8< --- */
if (type == LS_SF_DT_INTEGER)
{
prio->urgency = atoi(str);
if (prio->urgency <= 7)
prio->flags |= URG_SET;
else
{
printf("invalid value of urgency: %.*s\n", (int) len, str);
return -1;
}
/* --- 8< --- snip --- 8< --- */
}
And that's it! (For full HTTP Priority Properties example, see h3prio.c). The callback just keeps on getting called
for each data type until the callback returns a non-zero value, until the end of input, or until a parsing error occurs.
ls_sf_parse()
returns 0 on success and a negative value on failure.
Creation of in-memory structures to represent the parsed Structured Fields and base64-decoding of Byte Sequences are left as an exercise for the caller. There are two reasons for making this design choice:
- Simplicity. This parser can be used as a foundation for any Structured Fields representation of your choice.
- Efficiency. No reason to base64-decode something before knowing whether it is needed at all.
Please see some additional notes in the relatively short header file.
Just copy ls-sfparser.h and ls-sfparser.c into your source tree.
The last two arguments to ls_sf_parse()
may be used to pass a scratch space buffer for the parser to use. (Otherwise, it allocates
the memory it needs dynamically.) Now the parser does not call malloc(3)
at all!
ls_sf_parse()
returns -2 if a memory allocation error occurs. If the buffer you pass is too small, the parser just may run out.
This is a design feature.
This is a flex-generated parser with some modifications.
The post-generation alterations to the C source code have to do with memory management. The code has been changed to return
an error instead of calling yy_fatal_error()
, which is an awkward way to handle memory allocation failures.