Implement JSON framing function (Fixes #2)

3 files changed, 296 insertions, 1 deletions

diff --git a/plugins/framing_json.c b/plugins/framing_json.c
new file mode 100644
index 0000000..677e4e7
--- /dev/null
+++ b/plugins/framing_json.c
@@ -0,0 +1,279 @@
+#include <stdio.h>
+#include <ctype.h>
+#include <string.h>
+
+#include "../websocksy.h"
+
+static ssize_t json_length(char* data, size_t length);
+static ssize_t json_length_string(char* data, size_t length);
+
+static ssize_t json_length_array(char* data, size_t length){
+	size_t current_offset = 1;
+	int64_t next_length = 0;
+
+	if(!length){
+		return 0;
+	}
+
+	if(data[0] != '['){
+		return -1;
+	}
+
+	do{
+		//skip whitespace
+		for(; current_offset < length && isspace(data[current_offset]); current_offset++){
+		}
+
+		//empty array
+		if(!next_length && data[current_offset] == ']'){
+			break;
+		}
+
+		//get length
+		next_length = json_length(data + current_offset, length - current_offset);
+		if(next_length <= 0){
+			return next_length;
+		}
+
+		current_offset += next_length;
+		//check for end or ,
+		for(; current_offset < length && isspace(data[current_offset]); current_offset++){
+		}
+		//the ordering matters, check for complete array before checking for eod
+		if(data[current_offset] == ']'){
+			break;
+		}
+		if(current_offset >= length){
+			return 0;
+		}
+		if(data[current_offset] != ','){
+			break;
+		}
+		current_offset++;
+	} while(current_offset < length);
+
+	if(data[current_offset] == ']'){
+		return current_offset + 1;
+	}
+
+	return -1;
+}
+
+static ssize_t json_length_object(char* data, size_t length){
+	size_t current_offset = 1;
+	int64_t next_length = 0;
+
+	if(!length){
+		return 0;
+	}
+
+	if(data[0] != '{'){
+		return -1;
+	}
+
+	do{
+		//skip whitespace
+		for(; current_offset < length && isspace(data[current_offset]); current_offset++){
+		}
+
+		//empty object
+		if(!next_length && data[current_offset] == '}'){
+			break;
+		}
+
+		//get key string length
+		next_length = json_length_string(data + current_offset, length - current_offset);
+		if(next_length <= 0){
+			return next_length;
+		}
+
+		current_offset += next_length;
+		//find colon
+		for(; current_offset < length && isspace(data[current_offset]); current_offset++){
+		}
+		if(current_offset >= length - 1){ //need one more character
+			return 0;
+		}
+		else if(data[current_offset] != ':'){
+			return -1;
+		}
+		current_offset++;
+
+		//get value length
+		next_length = json_length(data + current_offset, length - current_offset);
+		if(next_length <= 0){
+			return next_length;
+		}
+
+		current_offset += next_length;
+		//check for end or ,
+		for(; current_offset < length && isspace(data[current_offset]); current_offset++){
+		}
+		//the ordering matters, check for complete object before checking for eod
+		if(data[current_offset] == '}'){
+			break;
+		}
+		if(current_offset >= length){
+			return 0;
+		}
+		if(data[current_offset] != ','){
+			break;
+		}
+		current_offset++;
+	} while(current_offset < length);
+
+	if(data[current_offset] == '}'){
+		return current_offset + 1;
+	}
+	return -1;
+}
+
+static ssize_t json_length_string(char* data, size_t length){
+	size_t string_length = 0;
+
+	if(!length){
+		return 0;
+	}
+
+	if(data[0] != '"'){
+		return -1;
+	}
+
+	//find terminating quotation mark not preceded by escape
+	for(string_length = 1; string_length < length
+				&& isprint(data[string_length])
+				&& (data[string_length] != '"' || data[string_length - 1] == '\\'); string_length++){
+	}
+
+	//complete string found
+	if(data[string_length] == '"' && data[string_length - 1] != '\\'){
+		return string_length + 1;
+	}
+
+	//still missing data
+	if(string_length >= length){
+		return 0;
+	}
+
+	//anything else
+	return -1;
+}
+
+static ssize_t json_length_value(char* data, size_t length){
+	size_t value_length = 0;
+
+	if(!length){
+		return 0;
+	}
+
+	if(data[0] == '-' || isdigit(data[0])){
+		//a number consisting of [minus] int [.int] [eE[+-]]
+		//terminator (comma, space, eod, eoa, eoo)
+		//since this is not a full parser, just ensure all characters are from that general set...
+		for(value_length = 1; value_length < length &&
+					(isdigit(data[value_length])
+					|| data[value_length] == '+'
+					|| data[value_length] == '-'
+					|| data[value_length] == '.'
+					|| tolower(data[value_length]) == 'e'); value_length++){
+		}
+		if(data[value_length] == ','
+				|| data[value_length] == ' '
+				|| data[value_length] == '}'
+				|| data[value_length] == ']'){
+			return value_length;
+		}
+		if(value_length >= length){
+			if(data[value_length - 1] == '.'
+					|| tolower(data[value_length - 1]) == 'e'
+					|| data[value_length - 1] == '-'
+					|| data[value_length - 1] == '+'){
+				return 0;
+			}
+			//numbers can be a value on their own...
+			return value_length;
+		}
+	}
+
+	//match complete values
+	if(length >= 4 && !strncmp(data, "null", 4)){
+		return 4;
+	}
+	else if(length >= 4 && !strncmp(data, "true", 4)){
+		return 4;
+	}
+	else if(length >= 5 && !strncmp(data, "false", 5)){
+		return 5;
+	}
+
+	//match prefix values
+	if(length < 4 && !strncmp(data, "null", length)){
+		return 0;
+	}
+	else if(length < 4 && !strncmp(data, "true", length)){
+		return 0;
+	}
+	else if(length < 5 && !strncmp(data, "false", length)){
+		return 0;
+	}
+
+	//invalid value
+	return -1;
+}
+
+static ssize_t json_length(char* data, size_t length){
+	size_t total_length = 0, current_offset = 0;
+	ssize_t rv = 0;
+
+	if(!length){
+		return 0;
+	}
+
+	//advance to the first non-blank
+	for(current_offset = 0; current_offset < length && isspace(data[current_offset]); current_offset++){
+	}
+
+	switch(data[current_offset]){
+		case '[':
+			rv = json_length_array(data + current_offset, length - current_offset);
+			break;
+		case '{':
+			rv = json_length_object(data + current_offset, length - current_offset);
+			break;
+		case '"':
+			rv = json_length_string(data + current_offset, length - current_offset);
+			break;
+		default:
+			//false null true
+			//0-9 | - -> number
+			rv = json_length_value(data + current_offset, length - current_offset);
+			break;
+	}
+
+	if(rv <= 0){
+		return rv;
+	}
+
+	return current_offset + rv;
+}
+
+static int64_t framing_json(uint8_t* data, size_t length, size_t last_read, ws_operation* opcode, void** framing_data, const char* config){
+	ssize_t data_length = json_length(data, length);
+
+	if(data_length > 0){
+		*opcode = ws_frame_text;
+		return data_length;
+	}
+	//incomplete
+	else if(data_length == 0){
+		return 0;
+	}
+	//failed to parse as json, just dump it
+	else{
+		return length;
+	}
+}
+
+static void __attribute__((constructor)) init(){
+	core_register_framing("json", framing_json);
+}
diff --git a/plugins/framing_json.md b/plugins/framing_json.md
new file mode 100644
index 0000000..c5a7e09
--- /dev/null
+++ b/plugins/framing_json.md
@@ -0,0 +1,16 @@
+# The `json` framing function
+
+The `json` framing function segments the peer stream at boundaries defined by complete JSON
+entities (Objects, Arrays, Strings or Values) and forwards these as text frames.
+
+For example, on encountering the start of a JSON object at the beginning of the stream, it will read from the
+peer until that objects closing brace and the send the entire object as a single text frame.
+
+The plugin does not implement a full parser and performs only rudimentary syntactic analysis to determine
+the end of the current entity. When encountering an error, the entire buffer is forwarded as a binary frame.
+It is possible to construct erroneous JSON objects that pass the parser without an error, however any
+syntactically correct object should be forwarded correctly (please file a bug with an example otherwise).
+
+## Configuration
+
+The `json` framing function does not require any configuration.
diff --git a/plugins/makefile b/plugins/makefile
index feab4cd..b09466d 100644
--- a/plugins/makefile
+++ b/plugins/makefile
@@ -1,5 +1,5 @@
 .PHONY: all clean
-PLUGINS = backend_file.so framing_fixedlength.so
+PLUGINS = backend_file.so framing_fixedlength.so framing_json.so
 
 CFLAGS += -fPIC -g -I../
 LDFLAGS += -shared