implement new "dashIsEmpty" parameter for string parser

Author: rgerhards

doc/configuration.rst

@@ -682,6 +682,16 @@ it becomes resolvable, and ``f`` will contain "12" and ``r`` will contain ":34 5
 This also shows the risk associated, as the result obtained may not necessarily be
 what was intended.
 
+option.dashIsEmpty
+~~~~~~~~~~~~~~~~~~
+
+This parameter, if True, permits to treat string consisting only of dash characters ('-') as
+being empty. The default value is False.
+
+This parameter is meant for e.g. processing web log data where a dash indicates a missing
+value but the user does not populate an analysis backend with dashes where "empty value" is
+meant.
+
 
 word
 ####

src/parser.c

@@ -1,6 +1,6 @@
 /*
  * liblognorm - a fast samples-based log normalization library
- * Copyright 2010-2018 by Rainer Gerhards and Adiscon GmbH.
+ * Copyright 2010-2021 by Rainer Gerhards and Adiscon GmbH.
  *
  * Modified by Pavel Levshin ([email protected]) in 2013
  *
@@ -3204,6 +3204,7 @@ struct data_String {
 		unsigned esc_md : 2;
 	} flags;
 	enum { ST_MATCH_EXACT = 0, ST_MATCH_LAZY = 1} matching;
+	int dashIsEmpty;
 	char qchar_begin;
 	char qchar_end;
 	char perm_chars[256]; // TODO: make this bit-wise, so we need  only 32 bytes
@@ -3373,6 +3374,14 @@ PARSER_Parse(String)
 	if(value != NULL) {
 		size_t strt;
 		size_t len;
+		if(data->dashIsEmpty) {
+			if(  (bHaveQuotes  && *parsed == 3 && !strncmp(npb->str+(*offs), "\"-\"", 3))
+			  || (!bHaveQuotes && *parsed == 1 && npb->str[*offs] == '-') ) {
+				*value = json_object_new_string_len("", 0);
+				r = 0;
+				goto done; /* shortcut exit */
+			}
+		}
 		if(bHaveQuotes && data->flags.strip_quotes) {
 			strt = *offs + 1;
 			len = *parsed - 2; /* del begin AND end quote! */
@@ -3498,6 +3507,8 @@ PARSER_Construct(String)
 				r = LN_BADCONFIG;
 				goto done;
 			}
+		} else if(!strcasecmp(key, "option.dashIsEmpty")) {
+			data->dashIsEmpty = json_object_get_boolean(val);
 		} else {
 			ln_errprintf(ctx, 0, "invalid param for hexnumber: %s",
 				 json_object_to_json_string(val));

tests/Makefile.am

@@ -67,6 +67,7 @@ TESTS_SHELLSCRIPTS = \
 	field_string_perm_chars.sh \
 	field_string_lazy_matching.sh \
 	field_string_doc_sample_lazy.sh \
+	field_string_dashIsEmpty.sh \
 	field_number.sh \
 	field_number-fmt_number.sh \
 	field_number_maxval.sh \

tests/field_string_dashIsEmpty.sh

@@ -0,0 +1,27 @@
+#!/bin/bash
+# added 2021-06-07 by Rainer Gerhards
+# This file is part of the liblognorm project, released under ASL 2.0
+. $srcdir/exec.sh
+no_solaris10
+test_def $0 "quoted string with dash"
+
+add_rule 'version=2'
+add_rule 'rule=:%
+		{"type":"string", "name":"str", "option.dashIsEmpty":True}
+	 %'
+
+execute '"-"'
+assert_output_json_eq '{ "str": ""}'
+
+reset_rules
+add_rule 'version=2'
+add_rule 'rule=:%
+		{"type":"quoted-string", "name":"str"}
+	 %'
+
+execute '"-"'
+assert_output_json_eq '{ "str": "\"-\""}'
+
+
+cleanup_tmp_files
+