Carry the municipality on a CEP that does not existHEADv0.2.0main

A CEP that fails is rarely a random number. It is usually someone who knows
the city and guessed the digits, and the classic guess is the -000 that 957
municipalities never received. The API now answers those 404s with a faixa
block naming the municipality that owns the band, so the library can say
"25845000 does not exist in Areal/RJ" instead of "not found".

It stays an error: lookup still throws Cepis::Error::CEPNotFound, and the
faixa hangs off the exception rather than turning a miss into an address.
Parsing it is deliberately forgiving — a missing, empty or malformed block
yields undef, because losing the clean not-found to a parse failure would be
a worse outcome than losing the context.

cep_ini and cep_fim are the smallest and largest CEP that actually exist,
not the bounds the Correios publish, and the interval between them is
sparse. That is why ceps is exposed alongside them and why the docs say
twice that the pair is not a list of valid CEPs.

8 files changed, 284 insertions, 18 deletions

diff --git a/lib/Cepis.pm b/lib/Cepis.pm
index 224d80d..1ea5b00 100644
--- a/lib/Cepis.pm
+++ b/lib/Cepis.pm
@@ -8,10 +8,11 @@ use Exporter 5.57 qw(import);
 use Cepis::Address;
 use Cepis::Client;
 use Cepis::Error;
+use Cepis::Faixa;
 use Cepis::Geo;
 use Cepis::Health;
 
-our $VERSION = '0.1.0';
+our $VERSION = '0.2.0';
 
 our @EXPORT_OK = qw(lookup lookup_many health normalize_cep);
 
@@ -101,6 +102,18 @@ the lot, and the split below it is by whether retrying helps:
         }
     }
 
+A C<CEPNotFound> often knows B<where> the CEP would have been. When the API says
+so, C<< $err->faixa >> is a L<Cepis::Faixa> naming the municipality that owns
+that band, and the message names it too:
+
+    # "CEP 25845000 does not exist in the DNE database (it would fall in Areal/RJ)"
+
+It is still a failure — C<faixa> is context for the error, never an address —
+and it is C<undef> when the CEP fell in a band no municipality owns. It is worth
+checking because a failing CEP is rarely random: it is usually someone who knows
+the city and guessed the number, and the classic guess is the C<-000> that 957
+municipalities never received.
+
 =head2 Field names
 
 The API surface is English, but the B<data fields keep the API's own Portuguese
@@ -169,7 +182,7 @@ The L<Cepis::Client> behind the functions above, built on first use.
 =head1 SEE ALSO
 
 L<Cepis::Client>, L<Cepis::Address>, L<Cepis::Geo>, L<Cepis::Health>,
-L<Cepis::Error>
+L<Cepis::Error>, L<Cepis::Faixa>
 
 The API itself is documented at L<https://cep.is>.
 
diff --git a/lib/Cepis/Address.pm b/lib/Cepis/Address.pm
index a938d6f..d1ae21f 100644
--- a/lib/Cepis/Address.pm
+++ b/lib/Cepis/Address.pm
@@ -5,7 +5,7 @@ use warnings;
 
 use Cepis::Geo;
 
-our $VERSION = '0.1.0';
+our $VERSION = '0.2.0';
 
 our @FIELDS = qw(
     cep logradouro complemento unidade bairro localidade
diff --git a/lib/Cepis/CSV.pm b/lib/Cepis/CSV.pm
index 6fabc0c..7ef6880 100644
--- a/lib/Cepis/CSV.pm
+++ b/lib/Cepis/CSV.pm
@@ -3,7 +3,7 @@ package Cepis::CSV;
 use strict;
 use warnings;
 
-our $VERSION = '0.1.0';
+our $VERSION = '0.2.0';
 
 sub parse_rows {
     my ($text) = @_;
diff --git a/lib/Cepis/Client.pm b/lib/Cepis/Client.pm
index 35eb51b..e75868b 100644
--- a/lib/Cepis/Client.pm
+++ b/lib/Cepis/Client.pm
@@ -13,9 +13,10 @@ use Time::HiRes ();
 use Cepis::Address;
 use Cepis::CSV;
 use Cepis::Error;
+use Cepis::Faixa;
 use Cepis::Health;
 
-our $VERSION = '0.1.0';
+our $VERSION = '0.2.0';
 
 use constant DEFAULT_URL => 'https://cep.is';
 use constant MAX_BATCH   => 500;
@@ -120,8 +121,10 @@ sub lookup {
     my ( $status, $body ) = $self->_request("/$number");
 
     if ( $status == 404 ) {
-        Cepis::Error::CEPNotFound->throw(
-            message => "CEP $number does not exist in the DNE database" );
+        my $faixa   = _faixa($body);
+        my $message = "CEP $number does not exist in the DNE database";
+        $message .= ' (it would fall in ' . $faixa->label . ')' if $faixa;
+        Cepis::Error::CEPNotFound->throw( message => $message, faixa => $faixa );
     }
     if ( $status == 400 ) {
         Cepis::Error::InvalidCEP->throw(
@@ -394,6 +397,14 @@ sub _error_message {
     return "$data->{mensagem}";
 }
 
+sub _faixa {
+    my ($body) = @_;
+
+    my $data = eval { $JSON->decode( defined $body ? $body : '' ) };
+    return undef unless ref $data eq 'HASH';
+    return Cepis::Faixa->from_json( $data->{faixa} );
+}
+
 sub _decode {
     my ($body) = @_;
     return '' unless defined $body;
@@ -575,6 +586,10 @@ A 400 from the API is surfaced as L<Cepis::Error::InvalidCEP> carrying the API's
 own message: the format already passed local validation, so the server knows
 something this client does not.
 
+A 404 becomes L<Cepis::Error::CEPNotFound>, and that error may carry a
+L<Cepis::Faixa> in C<< $err->faixa >> naming the municipality the CEP would have
+belonged to — see L</"A not-found that knows where it was">.
+
 Throws L<Cepis::Error::InvalidCEP> when the CEP is not 8 digits (no request is
 made), L<Cepis::Error::CEPNotFound> when it is well-formed but absent from the
 database, L<Cepis::Error::RateLimited> on 429 after every attempt,
@@ -583,6 +598,27 @@ L<Cepis::Error::Network> when the request never completed, and
 L<Cepis::Error::InvalidResponse> when a success status carries an unreadable
 body.
 
+=head3 A not-found that knows where it was
+
+    my $address = eval { $client->lookup('25845000') };
+    if ( !$address && ref $@ && $@->isa('Cepis::Error::CEPNotFound') ) {
+        if ( my $faixa = $@->faixa ) {
+            printf "%s does not exist in %s (%d CEPs, %s..%s)\n",
+                '25845000', $faixa->label, $faixa->ceps,
+                $faixa->cep_ini, $faixa->cep_fim;
+        }
+    }
+
+The API's 404 body may carry a C<faixa> block naming the municipality that owns
+the band the CEP fell into. When it does, the error carries a L<Cepis::Faixa>
+and its message names the municipality; when the CEP fell in a band nobody owns,
+the API omits the block and C<< $err->faixa >> is C<undef>.
+
+B<It is still a 404.> Nothing here turns a not-found into an address; the
+lookup threw, and C<faixa> only says where the CEP would have been. Nor is
+C<cep_ini>..C<cep_fim> a list of valid CEPs — the range is sparse, which is what
+C<ceps> is for. L<Cepis::Faixa> spells both out.
+
 =head2 lookup_many
 
     my $result = $client->lookup_many( \@ceps );
@@ -643,7 +679,7 @@ C<500>, the most CEPs the API accepts in a single batch request.
 
 =head1 SEE ALSO
 
-L<Cepis>, L<Cepis::Address>, L<Cepis::Health>, L<Cepis::Error>
+L<Cepis>, L<Cepis::Address>, L<Cepis::Health>, L<Cepis::Error>, L<Cepis::Faixa>
 
 =head1 LICENCE
 
diff --git a/lib/Cepis/Error.pm b/lib/Cepis/Error.pm
index 080b0a1..37f2a08 100644
--- a/lib/Cepis/Error.pm
+++ b/lib/Cepis/Error.pm
@@ -3,7 +3,7 @@ package Cepis::Error;
 use strict;
 use warnings;
 
-our $VERSION = '0.1.0';
+our $VERSION = '0.2.0';
 
 use overload
     '""'     => sub { $_[0]->message },
@@ -30,7 +30,7 @@ use strict;
 use warnings;
 use parent -norequire, 'Cepis::Error';
 
-our $VERSION = '0.1.0';
+our $VERSION = '0.2.0';
 
 package Cepis::Error::CEPNotFound;
 
@@ -38,7 +38,9 @@ use strict;
 use warnings;
 use parent -norequire, 'Cepis::Error';
 
-our $VERSION = '0.1.0';
+our $VERSION = '0.2.0';
+
+sub faixa { return $_[0]->{faixa} }
 
 package Cepis::Error::Network;
 
@@ -46,7 +48,7 @@ use strict;
 use warnings;
 use parent -norequire, 'Cepis::Error';
 
-our $VERSION = '0.1.0';
+our $VERSION = '0.2.0';
 
 package Cepis::Error::ServerError;
 
@@ -54,7 +56,7 @@ use strict;
 use warnings;
 use parent -norequire, 'Cepis::Error';
 
-our $VERSION = '0.1.0';
+our $VERSION = '0.2.0';
 
 sub status { return $_[0]->{status} }
 
@@ -64,7 +66,7 @@ use strict;
 use warnings;
 use parent -norequire, 'Cepis::Error';
 
-our $VERSION = '0.1.0';
+our $VERSION = '0.2.0';
 
 sub retry_after { return $_[0]->{retry_after} }
 
@@ -74,7 +76,7 @@ use strict;
 use warnings;
 use parent -norequire, 'Cepis::Error';
 
-our $VERSION = '0.1.0';
+our $VERSION = '0.2.0';
 
 1;
 
@@ -171,6 +173,25 @@ message.
 
 The CEP is well-formed but does not exist in the DNE database.
 
+=over 4
+
+=item C<< $err->faixa >>
+
+A L<Cepis::Faixa> naming the municipality the CEP I<would> belong to, or
+C<undef> when it fell in a band no municipality owns. The API omits the block
+rather than sending null, and a malformed one is dropped rather than turning the
+not-found into a parse failure.
+
+It is context for the error, B<not> an address: the lookup still failed. What it
+buys is telling the user "25845000 does not exist in Areal/RJ" instead of "not
+found" — worth having, because a failing CEP is rarely random. It is usually
+someone who knows the city and guessed the number, and the classic guess is the
+C<-000> that 957 municipalities never received.
+
+When it is present, the message already names the municipality.
+
+=back
+
 =head2 Cepis::Error::Network
 
 The request never completed: DNS, refused connection, timeout, TLS.
@@ -217,7 +238,7 @@ as a diagnosis: the problem is not in your code.
 
 =head1 SEE ALSO
 
-L<Cepis>, L<Cepis::Client>
+L<Cepis>, L<Cepis::Client>, L<Cepis::Faixa>
 
 =head1 LICENCE
 
diff --git a/lib/Cepis/Faixa.pm b/lib/Cepis/Faixa.pm
new file mode 100644
index 0000000..7dabdc3
--- /dev/null
+++ b/lib/Cepis/Faixa.pm
@@ -0,0 +1,196 @@
+package Cepis::Faixa;
+
+use strict;
+use warnings;
+
+use Scalar::Util ();
+
+our $VERSION = '0.2.0';
+
+sub from_json {
+    my ( $class, $data ) = @_;
+
+    return undef unless ref $data eq 'HASH';
+
+    for my $key (qw(localidade cep_ini cep_fim ceps)) {
+        return undef unless defined $data->{$key};
+        return undef if ref $data->{$key};
+    }
+    return undef unless length "$data->{localidade}";
+    for my $key (qw(cep_ini cep_fim)) {
+        return undef unless "$data->{$key}" =~ /\A[0-9]{8}\z/;
+    }
+    return undef unless Scalar::Util::looks_like_number( $data->{ceps} );
+
+    return bless {
+        localidade => "$data->{localidade}",
+        uf         => _string( $data->{uf} ),
+        estado     => _string( $data->{estado} ),
+        regiao     => _string( $data->{regiao} ),
+        ibge       => _string( $data->{ibge} ),
+        cep_ini    => "$data->{cep_ini}",
+        cep_fim    => "$data->{cep_fim}",
+        ceps       => int( 0 + $data->{ceps} ),
+    }, $class;
+}
+
+sub localidade { return $_[0]->{localidade} }
+sub uf         { return $_[0]->{uf} }
+sub estado     { return $_[0]->{estado} }
+sub regiao     { return $_[0]->{regiao} }
+sub ibge       { return $_[0]->{ibge} }
+sub cep_ini    { return $_[0]->{cep_ini} }
+sub cep_fim    { return $_[0]->{cep_fim} }
+sub ceps       { return $_[0]->{ceps} }
+
+sub label {
+    my ($self) = @_;
+    return length $self->{uf}
+        ? "$self->{localidade}/$self->{uf}"
+        : $self->{localidade};
+}
+
+sub to_hash {
+    my ($self) = @_;
+    return {
+        localidade => $self->{localidade},
+        uf         => $self->{uf},
+        estado     => $self->{estado},
+        regiao     => $self->{regiao},
+        ibge       => $self->{ibge},
+        cep_ini    => $self->{cep_ini},
+        cep_fim    => $self->{cep_fim},
+        ceps       => $self->{ceps},
+    };
+}
+
+sub _string {
+    my ($value) = @_;
+    return '' unless defined $value;
+    return '' if ref $value;
+    return "$value";
+}
+
+1;
+
+__END__
+
+=encoding UTF-8
+
+=head1 NAME
+
+Cepis::Faixa - the municipality a CEP that does not exist would belong to
+
+=head1 SYNOPSIS
+
+    my $address = eval { Cepis::lookup('25845000') };
+    if ( !$address && ref $@ && $@->isa('Cepis::Error::CEPNotFound') ) {
+
+        if ( my $faixa = $@->faixa ) {
+            print $faixa->localidade, "\n";    # "Areal"
+            print $faixa->uf,         "\n";    # "RJ"
+            print $faixa->ceps,       "\n";    # 118
+            printf "%s..%s\n", $faixa->cep_ini, $faixa->cep_fim;
+            # 25845001..25849899
+        }
+    }
+
+=head1 DESCRIPTION
+
+B<This is context for an error, not an address.> The lookup still failed, the
+CEP still does not exist, and L<Cepis::Client/lookup> still throws
+L<Cepis::Error::CEPNotFound>. What this object adds is I<which municipality
+owns the band the CEP fell into> — enough to tell the user "25845000 does not
+exist in Areal/RJ" instead of "not found".
+
+That matters because a failing CEP is rarely random: it is usually someone who
+knows the city and guessed the number, and the classic guess is the C<-000>
+that 957 municipalities never received.
+
+The object is reachable as C<< $err->faixa >> on L<Cepis::Error::CEPNotFound>,
+and is C<undef> whenever the API sent no C<faixa> — the API B<omits> the field
+rather than sending null when the CEP falls in a band no municipality owns.
+
+=head2 The range is not a list of valid CEPs
+
+L</cep_ini> and L</cep_fim> are the B<smallest and the largest CEP that actually
+exist> in that municipality. They are not the bounds the Correios publish: the
+DNE's own start for Areal is C<25845000> — exactly the CEP that just failed.
+
+The range is also B<sparse>. Areal's 118 CEPs are scattered across the roughly
+4,900 numbers between C<25845001> and C<25849899>. Iterating the interval, or
+presenting it as "the valid CEPs for this city", produces mostly numbers that do
+not exist. L</ceps> is there precisely so that cannot pass unnoticed: it says how
+many of them are real.
+
+Field names are the API's own, in Portuguese. See L<Cepis::Address> for why.
+
+=head1 CONSTRUCTOR
+
+=head2 from_json
+
+    my $faixa = Cepis::Faixa->from_json( $hashref );
+
+Builds from the C<faixa> block of a 404 body. Returns C<undef> when the block is
+missing, empty, not a hash, or malformed.
+
+Parsing here is deliberately forgiving: a bad C<faixa> must never turn a clean
+not-found into a parse failure. The caller still gets its
+L<Cepis::Error::CEPNotFound>, just without the extra context.
+
+=head1 ACCESSORS
+
+=head2 localidade
+
+Municipality name. Always present and non-empty — a block without it does not
+build.
+
+=head2 uf
+
+Two-letter state code.
+
+=head2 estado
+
+Full state name.
+
+=head2 regiao
+
+Region of the country, such as C<"Sudeste">.
+
+=head2 ibge
+
+IBGE municipality code. The stable way to identify the municipality; the name
+alone is not unique across states.
+
+=head2 cep_ini
+
+The smallest CEP that exists in the municipality, unmasked. Read the warning
+above before treating it as a lower bound.
+
+=head2 cep_fim
+
+The largest CEP that exists in the municipality, unmasked.
+
+=head2 ceps
+
+How many CEPs the municipality actually has, as an integer. Compare it with the
+span between L</cep_ini> and L</cep_fim> to see how sparse the range is.
+
+=head2 label
+
+C<"Areal/RJ">: the municipality and its state, ready to put in a message. Falls
+back to the municipality alone when the API sent no C<uf>.
+
+=head2 to_hash
+
+The eight fields as a plain hashref.
+
+=head1 SEE ALSO
+
+L<Cepis>, L<Cepis::Error>, L<Cepis::Client>
+
+=head1 LICENCE
+
+ISC. Copyright (c) 2026 Murilo Ijanc E<lt>[email protected]<gt>.
+
+=cut
diff --git a/lib/Cepis/Geo.pm b/lib/Cepis/Geo.pm
index bee2515..c2c8fa3 100644
--- a/lib/Cepis/Geo.pm
+++ b/lib/Cepis/Geo.pm
@@ -5,7 +5,7 @@ use warnings;
 
 use Scalar::Util ();
 
-our $VERSION = '0.1.0';
+our $VERSION = '0.2.0';
 
 sub from_json {
     my ( $class, $data ) = @_;
diff --git a/lib/Cepis/Health.pm b/lib/Cepis/Health.pm
index 6782fd0..858b543 100644
--- a/lib/Cepis/Health.pm
+++ b/lib/Cepis/Health.pm
@@ -5,7 +5,7 @@ use warnings;
 
 use Scalar::Util ();
 
-our $VERSION = '0.1.0';
+our $VERSION = '0.2.0';
 
 sub from_json {
     my ( $class, $data ) = @_;