Return lineno:column alongside error messages

[oisanjaya]

This PR adds additioinal line number and column information on error messages.

What I've done in this PR was declaring static int variable in parser.c named current_line_number and current_column_number, Theese variables value initialized with 1.

Every kdl_parser_next_event() loop I increase value of current_column_number.

Every time parser find KDL_TOKEN_NEWLINE, current_line_number is increased and current_column_number reset back to 1.

Then finally inside _set_parse_error() I allocate new string then format it using snprintf prepending current_line_number current_column_number into final error message

[tjol]

I'm guessing this will give incorrect line numbers after any semicolon

[tjol]

Hi! Thanks for giving this a go!

Without having tested it, I would assume that this sometimes gives incorrect line numbers (particularly when there are semicolons involved) and usually gives incorrect column numbers (since the code appears to assume that 1 token = 1 character, which is normally not true)

I think that the tokenizer would have to keep track of (and return) the line and column for each token for the numbers to be correct.

[oisanjaya]

OK I'll admit my last PR was a half assed effort. Here's my updated PR.

current_column_number and current_line_number initialized to 1 in kdl_create_string_tokenizer() and kdl_create_stream_tokenizer()

current_column_number increased every _tok_get_char() call

current_column_number decreased back when:

• _remove_inital_bom() call
• kdl_pop_token() loop when:
  • End of find whitespace run (overshoot, traceback one character)
  • Current character will be consumed by _pop_string() i.e. found ", r, or #
  • _pop_string() returns KDL_TOKENIZER_OK (overshoot, traceback one character)
  • Current character will be consumed by _pop_word()
• _pop_word() when end_of_word reached (overshoot, traceback one character)
• Inside _pop_string():
  • After make determine raw/hash/regular string
  • After counting hash
  • After counting quotes

current_column_number reset to 1 and current_line_number increased upon finding newline:

• Inside kdl_pop_token() loop
• Inside _pop_string()

kdl_token struct now includes line_no and col_no to store token location

_set_parse_error() of parser.c now expects current token as an argument to build error message with problematic token location

Reverting back last PR:

• Remove current_column_number and current_line_number from parser.c as it is now handled in tokenizer.c

I feel like i still miss something, feel free to point me out