Errors with large data structures became harder to read

[oakes]

I use expound in edna and I noticed that errors in 0.7.1 were harder to read than in 0.7.0 when they occur inside a large outer data structure. In 0.7.0, it points to the specific part of the data structure first:

Exception in thread "main" java.lang.Exception: -- Syntax error -------------------

  [...
   [:guitarx
    ^^^^^^^^
    ...
    ...
    ...
    ...
    ...
    ...
    ...
    ...
    ...
    ...
    ...
    ...]
<snip>

In 0.7.1, it prints the entire data structure first, and you have to scroll (sometimes far) down to see it eventually point to the specific part of the data structure where the error is:

Exception in thread "main" java.lang.Exception: -- Spec failed --------------------

  [{:tempo 80}
   [:guitarx {:octave 3} 1/16 :b :+c 1/8 :+d :b :+c :a :b :g :a]
   [:banjo {:octave 3} 1/16 :b :+c 1/8 :+d :b :+c :a :b :g :a]
   [:guitar {:octave 3} 1/16 :b :+c 1/8 :+d :b :+c :a :b :g :a]
   [:guitar {:octave 3} 1/2 :d 1/8 :g :g :a :b :g :b 1/2 :a]
   [:banjo {:octave 3} 1/8 :g :g :a :b 1/2 :g]
   [:guitar {:octave 2} 1/16 :g :g 1/8 :g :a :b :+c :+d :+c 1/2 :b]
   [:banjo {:octave 3} 1/16 :g :g 1/8 :g :a :b :+c :+d :+c 1/2 :b]
   [:guitar {:octave 2} 1/16 :g :g 1/8 :g :a :b :+c :+d :+c 1/2 :b]
   [:banjo {:octave 3} 1/16 :g :g 1/8 :g :a :b :+c :+d :+c 1/2 :b]
   [:guitar {:octave 3} 1/16 :g :g 1/8 :g :a :b :+c :+d :+c 1/2 :b]
   [:banjo {:octave 3} 1/16 :g :g 1/8 :g :a :b :+c :+d :+c 1/2 :b]
   [:guitar
    {:octave 4}
    1/16
    #{:-g :-b :d}
    #{:-g :-b :d}
<snip>
-- Spec failed --------------------

  [...
   [:guitarx {:octave 3} 1/16 :b :+c 1/8 :+d :b :+c :a :b :g :a]
   ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
   ...
   ...
   ...
   ...
   ...
   ...
   ...
   ...
<snip>
-- Syntax error -------------------

  [...
   [:guitarx
    ^^^^^^^^
    ...
    ...
    ...
    ...
    ...
    ...
    ...
    ...
    ...
    ...
    ...
    ...]
<snip>

I posted a code example and the output of both versions here https://gist.github.com/oakes/dbc902e8fb077d6430fa0909cf4d7272

[bhb]

Thanks for reporting this! In 0.7.0, some relevant errors were omitted when there were multiple ways to fix the value (which occurred in macro specs which had many ‘or’ or ‘alt’ specs) so in 0.7.1, I changed the code to show every problem.

However, I agree this choice is quite verbose in this case where only the inner spec is relevant. I’ll think more about how to do the best thing in both cases.