Trim leading whitespace from contextual error, detect EOF

Side-note that the token location tuple maintains the correct column
number to include the whitespace in the source file. Side-side note:
tabs are counted as one column but many editors will count them as
whatever the tabstop/tabwidth is.

1 files changed, 7 insertions, 0 deletions

diff --git a/lexer.c b/lexer.c
index a8b7716..d9810a5 100644
--- a/lexer.c
+++ b/lexer.c
@@ -53,6 +53,7 @@ static char buf[BUFFER_SIZE];
 static FILE* fd;
 static size_t line_number = 0;
 static size_t column_number = 0;
+static size_t leading_whitespace_len = 0;
 static struct token *tok_start = NULL;
 static struct token *tok_cursor = NULL;
 
@@ -64,6 +65,7 @@ get_current_loc(void) {
 	struct location l;
 	l.line = line_number;
 	l.column = column_number + 1;
+	l.leading_whitespace_len = leading_whitespace_len;
 	return l;
 }
 
@@ -187,6 +189,8 @@ lex_colon(void) {
 static int
 lex_line(void) {
 	size_t length = strlen(buf);
+	leading_whitespace_len = 0;
+
 	while (column_number < length && column_number < strlen(buf)) {
 		switch (buf[column_number]) {
 			case ':':
@@ -195,6 +199,9 @@ lex_line(void) {
 			case ' ':
 			case '\t':
 				eat_whitespace();
+				if (leading_whitespace_len == 0) {
+					leading_whitespace_len = column_number;
+				}
 				break;
 			case '\r':
 			case '\n':