Fix Query Parser for Automated Transactions (#1)

* Add failing test for use case

TBH I don't know what I'm doing here, but this seems to fail for the
right reasons enough to reflect the parser bug here.

* Append to the ident on a closing brace ')'

When parsing the automated rule, a scanner reads the line left-to-right
char-by-char. The default behaviour is to append the char under the
cursor to some `ident` string. When the cursor is on a ')', it skips the
default handling and switches into some special handling: it tries to
test the string it's reading if it's one of the keywords it knows, to
select which type of token just got scanned. If what was scanned is not
a known token, it defaults to `token_t::TERM` and returns a new token
with the currently accumulated `ident` as a `token_t::TERM`. Issue is,
since it skipped the appending to do some custom handling, the
`token_t::TERM` will always be without its closing brace. The scanner
needs to append the character under the cursor if it's falling through
to default processing.

* fix test case

- ensure proper spacing for the posting to have an amount
- ensure the posting balances against an account
- the meaning of the number after `->` is the exit code

* undo wrong approach

* consume_next if unbalanced_braces

* how this can be extended

1 files changed, 2 insertions, 0 deletions

diff --git a/src/query.h b/src/query.h
index 20a7bc18..581c7533 100644
--- a/src/query.h
+++ b/src/query.h
@@ -231,6 +231,8 @@ public:
         token_cache = next_token(tok_context);
       return token_cache;
     }
+
+    bool unbalanced_braces(const string str);
   };
 
   enum kind_t {