Parse quotes and candidates for expansion/substitution as a tree

[emersion]

That means, don't unquote when parsing. That probably means: keep a list of {unquoted, quoted} fragments when parsing.

[emersion]

Err, it's probably more a tree than a flat list. A token can contain:

• Unquoted strings
• Quoted strings (" or ')
• Candidates for parameter expansion (${expression} or $variable)
• Candidates for command substitution ($(command) or `command`)
• Candidates for arithmetic expansion ($((expression)))

Examples:

echo "hey 'there'" # hey 'there'
echo 'hey "there"' # hey "there"
echo "hey $(echo "there")" # hey there
echo `echo \`echo hey\`` # hey
echo $(echo $(echo hey `echo there $(echo how are you)`)) # hey there how are you
echo $(($x-1))
echo ${x:-$(ls)} # ls is executed only if x is null or unset

WIP proposal:

enum mrsh_token_type {
    TOKEN_STRING,
    TOKEN_PARAMETER,
    TOKEN_COMMAND,
    TOKEN_ARITHMETIC,
    TOKEN_LIST,
};

struct mrsh_token {
    enum mrsh_token_type type;
};

struct mrsh_token_string {
    struct mrsh_token token;
    char *str;
    bool single_quoted;
};

struct mrsh_token_parameter {
    struct mrsh_token token;
    char *name;
    char *op;
    struct mrsh_token *arg;
};

struct mrsh_token_command {
    struct mrsh_token token;
    char *command;
    bool back_quoted;
};

struct mrsh_token_arithmetic {
    struct mrsh_token token;
    struct mrsh_token *expression;
};

struct mrsh_token_list {
    struct mrsh_token token;
    struct mrsh_array children; // struct mrsh_token *
    bool double_quoted;
};