tojsonl: improved boolean inferencing

by turning on case-insensitive frequency compilation, we can now properly infer if a field is boolean even if the domain is more than 2 case-sensitive, but 2 case-insensitive
(e.g. True, False, true, false, TRUE, FALSE, truE, fAlse has a case-sensitive domain of 8 and cardinality 8, but a case insensitive domain of 2, cardinality 2)

src/cmd/tojsonl.rs

@@ -112,6 +112,11 @@ pub fn run(argv: &[&str]) -> CliResult<()> {
         // is just two values. if its more than 2, that's all we need know
         // for boolean inferencing
         flag_enum_threshold:  3,
+        // ignore case for enum constraints
+        // so we can properly infer booleans. e.g. if a field has a domain of
+        // True, False, true, false, TRUE, FALSE that it is still a boolean
+        // with a case-insensitive cardinality of 2
+        flag_ignore_case:     true,
         flag_strict_dates:    false,
         flag_pattern_columns: crate::select::SelectColumns::parse("")?,
         // json doesn't have a date type, so don't infer dates

tests/test_tojsonl.rs

@@ -164,12 +164,13 @@ fn tojsonl_not_boolean_case_sensitive() {
     let mut cmd = wrk.command("tojsonl");
     cmd.arg("in.csv");
 
-    // not treated as boolean since col1's domain has three values
+    // properly treated as boolean since col1's domain has two values
+    // case-insensitive, even though the enum for col1 is
     // True, False and false
     let got: String = wrk.stdout(&mut cmd);
-    let expected = r#"{"col1":"True","col2":"Mark"}
-{"col1":"False","col2":"John"}
-{"col1":"false","col2":"Bob"}"#;
+    let expected = r#"{"col1":true,"col2":"Mark"}
+{"col1":false,"col2":"John"}
+{"col1":false,"col2":"Bob"}"#;
     assert_eq!(got, expected);
 }