Name the municipality behind a CEP that does not existHEADv0.2.0main

The API now hangs an optional faixa block off the 404 of GET /{cep}, and
this exposes it as CepNotFoundException.Faixa. A CEP that fails is rarely
random: it is usually someone who knows the city and guessed the number,
most often the -000 that 957 municipalities never received. Knowing where
the guess landed turns a dead-end "CEP not found" into a question worth
asking the user.

It stays an error. The block names the municipality the number would
belong to, never an address, and the lookup still throws. CepIni and
CepFim are the smallest and largest CEP that actually exist there, not the
range the Correios publish — the published start for Areal is 25845000,
exactly the CEP that fails — and the interval between them is sparse,
which is what Ceps is for.

Parsing is deliberately toothless: absent, misshapen or nameless, the
block becomes null and the caller gets the same clean not-found it got
before the field existed. A courtesy on top of a 404 must never be able to
turn it into a parse failure.

1 files changed, 67 insertions, 3 deletions

diff --git a/src/Cepis/CepisClient.cs b/src/Cepis/CepisClient.cs
index 72637c0..e32b78e 100644
--- a/src/Cepis/CepisClient.cs
+++ b/src/Cepis/CepisClient.cs
@@ -54,7 +54,7 @@ public sealed class CepisClient : IDisposable
     public const int MaxBatch = 500;
 
     /// <summary>The version of this library.</summary>
-    public const string Version = "0.1.0";
+    public const string Version = "0.2.0";
 
     private const string AcceptHeader = "application/json, text/csv;q=0.9, */*;q=0.1";
 
@@ -175,9 +175,17 @@ public sealed class CepisClient : IDisposable
     /// Look up a single CEP.
     /// </summary>
     /// <remarks>
+    /// <para>
     /// This is the only call that carries a coordinate. About 57% of CEPs have
     /// one; when they do not, <see cref="Address.Geo"/> is
     /// <see langword="null"/>.
+    /// </para>
+    /// <para>
+    /// A CEP that does not exist throws, as it always has. What is new is that
+    /// the exception may name the municipality the number would belong to, in
+    /// <see cref="CepNotFoundException.Faixa"/> — context for the failure, not
+    /// an address, and never a reason to treat the lookup as having succeeded.
+    /// </para>
     /// </remarks>
     /// <param name="cep">the CEP, masked or not</param>
     /// <param name="cancellationToken">cancels the call, waits included</param>
@@ -186,7 +194,10 @@ public sealed class CepisClient : IDisposable
     /// the CEP is not eight digits, in which case no request is made, or the API
     /// rejected it with a 400
     /// </exception>
-    /// <exception cref="CepNotFoundException">well-formed, but not in the database</exception>
+    /// <exception cref="CepNotFoundException">
+    /// well-formed, but not in the database; carries the municipality it would
+    /// belong to when the API names one
+    /// </exception>
     /// <exception cref="RateLimitedException">the API answered 429 on every attempt</exception>
     /// <exception cref="ServerErrorException">the API answered 5xx on every attempt</exception>
     /// <exception cref="NetworkException">the request never completed</exception>
@@ -198,7 +209,8 @@ public sealed class CepisClient : IDisposable
 
         if (reply.Status == 404)
         {
-            throw new CepNotFoundException($"CEP {number} does not exist in the DNE database");
+            var faixa = ReadFaixa(reply.Body);
+            throw new CepNotFoundException(NotFoundMessage(number, faixa), faixa);
         }
 
         if (reply.Status == 400)
@@ -459,6 +471,58 @@ public sealed class CepisClient : IDisposable
         return fallback;
     }
 
+    /// <summary>
+    /// Read the <c>faixa</c> block out of a 404 body, when it has one.
+    /// </summary>
+    /// <remarks>
+    /// The block is optional and always will be: the API omits it — it does not
+    /// send a null — for a CEP in a band no municipality owns. Absent,
+    /// misshapen or unreadable, the answer here is <see langword="null"/> and
+    /// the caller still gets the plain not-found it was going to get anyway.
+    /// </remarks>
+    /// <param name="body">the 404 body</param>
+    /// <returns>the municipality the CEP would belong to, or null</returns>
+    private static Faixa? ReadFaixa(byte[] body)
+    {
+        try
+        {
+            using var document = JsonDocument.Parse(new ReadOnlyMemory<byte>(body));
+            if (document.RootElement.ValueKind == JsonValueKind.Object
+                && document.RootElement.TryGetProperty("faixa", out var faixa))
+            {
+                return FaixaReader.Read(faixa);
+            }
+        }
+        catch (JsonException)
+        {
+            return null;
+        }
+
+        return null;
+    }
+
+    /// <summary>
+    /// What to tell the caller about a CEP that is not in the database.
+    /// </summary>
+    /// <remarks>
+    /// The municipality goes in the message when the API named one, because the
+    /// message is what ends up in a log, and "the range belongs to Areal/RJ" is
+    /// the difference between a dead end and a question worth asking the user.
+    /// The count is in it for the same reason the API sends it: so nobody reads
+    /// the two ends as a list of CEPs to walk.
+    /// </remarks>
+    /// <param name="cep">the eight digits</param>
+    /// <param name="faixa">the municipality it would belong to, if any</param>
+    /// <returns>the message</returns>
+    private static string NotFoundMessage(string cep, Faixa? faixa)
+    {
+        var message = $"CEP {cep} does not exist in the DNE database";
+        return faixa is null
+            ? message
+            : message + $" — the range belongs to {faixa.Localidade}/{faixa.UF}, "
+                + $"which has {faixa.Ceps} CEP(s) between {faixa.CepIni} and {faixa.CepFim}";
+    }
+
     private static string Preview(byte[] body) =>
         Encoding.UTF8.GetString(body, 0, Math.Min(body.Length, 120));