use Carp;
use IO::File;
use Params::Validate qw(:all);
-use List::MoreUtils qw(all pairwise);
+use List::MoreUtils qw(all pairwise firstidx);
use Text::CSV_XS;
use Rose::Object::MakeMethods::Generic scalar => [ qw(
file encoding sep_char quote_char escape_char header profile
numberformat dateformat ignore_unknown_columns strict_profile is_multiplexed
_row_header _io _csv _objects _parsed _data _errors all_cvar_configs case_insensitive_header
+ _multiplex_datatype_position
) ];
use SL::Helper::Csv::Dispatcher;
$self->_open_file;
return if ! $self->_check_multiplexed;
return if ! $self->_check_header;
+ return if ! $self->_check_multiplex_datatype_position;
return if ! $self->dispatcher->parse_profile;
return if ! $self->_parse_data;
croak 'must parse first' unless $self->_parsed;
$self->_make_objects unless $self->_objects;
- return wantarray ? @{ $self->_objects } : $self->_objects;
+ return $self->_objects;
}
sub errors {
# Each profile needs a class and a row_ident
my $info_ok = all { defined $_->{class} && defined $_->{row_ident} } @profile;
$self->_push_error([
+ undef,
0,
"missing class or row_ident in one of the profiles for multiplexed data",
0,
0]) unless $info_ok;
- # If header is given, there need to be a header for each profile
+ # If header is given, there needs to be a header for each profile
# and no empty headers.
if ($info_ok && $self->header) {
my @header = @{ $self->header };
my $t_ok = scalar @profile == scalar @header;
$self->_push_error([
+ undef,
0,
"number of headers and number of profiles must be the same for multiplexed data",
0,
$t_ok = all { scalar @$_ > 0} @header;
$self->_push_error([
+ undef,
0,
"no empty headers are allowed for multiplexed data",
0,
foreach my $p_num (0..$n_header - 1) {
my $h = $self->_csv->getline($self->_io);
+ my ($code, $string, $position, $record, $field) = $self->_csv->error_diag;
+
$self->_push_error([
$self->_csv->error_input,
- $self->_csv->error_diag,
- 0,
+ $code, $string, $position, $record // 0,
]) unless $h;
if ($self->is_multiplexed) {
# people insist that case sensitivity doesn't exist and try to enter all
# sorts of stuff. at this point we've got a profile (with keys that represent
# valid methods), and a header full of strings. if two of them match, the user
- # mopst likely meant that field, so rewrite the header
+ # most likely meant that field, so rewrite the header
if ($self->case_insensitive_header) {
die 'case_insensitive_header is only possible with profile' unless $self->profile;
if ($header) {
my $h_aref = ($self->is_multiplexed)? $header : [ $header ];
my $p_num = 0;
foreach my $h (@{ $h_aref }) {
- my @names = (
- keys %{ $self->profile->[$p_num]->{profile} || {} },
+ my %names = (
+ (map { $_ => $_ } keys %{ $self->profile->[$p_num]->{profile} || {} }),
+ (map { $_ => $self->profile->[$p_num]{mapping}{$_} } keys %{ $self->profile->[$p_num]->{mapping} || {} }),
);
- for my $name (@names) {
+ for my $name (keys %names) {
for my $i (0..$#$h) {
- $h->[$i] = $name if lc $h->[$i] eq lc $name;
+ $h->[$i] = $names{$name} if lc $h->[$i] eq lc $name;
}
}
$p_num++;
return $self->header($header);
}
+sub _check_multiplex_datatype_position {
+ my ($self) = @_;
+
+ return 1 if !$self->is_multiplexed; # ok if not multiplexed
+
+ my @positions = map { firstidx { 'datatype' eq lc($_) } @{ $_ } } @{ $self->header };
+ my $first_pos = $positions[0];
+ if (all { $first_pos == $_ } @positions) {
+ $self->_multiplex_datatype_position($first_pos);
+ return 1;
+ } else {
+ $self->_push_error([undef,
+ 0,
+ "datatype field must be at the same position for all datatypes for multiplexed data",
+ 0,
+ 0]);
+ return 0;
+ }
+}
+
+sub _is_empty_row {
+ return !!all { !$_ } @{$_[0]};
+}
+
sub _parse_data {
my ($self, %params) = @_;
my (@data, @errors);
while (1) {
my $row = $self->_csv->getline($self->_io);
if ($row) {
+ next if _is_empty_row($row);
my $header = $self->_header_by_row($row);
+ if (!$header) {
+ push @errors, [
+ undef,
+ 0,
+ "Cannot get header for row. Maybe row name and datatype field not matching.",
+ 0,
+ 0];
+ last;
+ }
my %hr;
@hr{@{ $header }} = @$row;
push @data, \%hr;
} else {
last if $self->_csv->eof;
+
# Text::CSV_XS 0.89 added record number to error_diag
- if (qv(Text::CSV_XS->VERSION) >= qv('0.89')) {
- push @errors, [
- $self->_csv->error_input,
- $self->_csv->error_diag,
- ];
- } else {
- push @errors, [
- $self->_csv->error_input,
- $self->_csv->error_diag,
- $self->_io->input_line_number,
- ];
- }
+ my ($code, $string, $position, $record, $field) = $self->_csv->error_diag;
+
+ push @errors, [
+ $self->_csv->error_input,
+ $code, $string, $position,
+ $record // $self->_io->input_line_number,
+ ];
}
last if $self->_csv->eof;
}
}
if ($self->is_multiplexed) {
- return $self->_row_header->{$row->[0]}
+ return $self->_row_header->{$row->[$self->_multiplex_datatype_position]}
} else {
return $self->header;
}
$self->_errors(\@new_errors);
}
+sub specs {
+ $_[0]->dispatcher->_specs
+}
1;
my $status = $csv->parse;
my $hrefs = $csv->get_data;
- my @objects = $csv->get_objects;
+ my $objects = $csv->get_objects;
my @errors = $csv->errors;
See Synopsis.
-Text::CSV offeres already good functions to get lines out of a csv file, but in
-most cases you will want those line to be parsed into hashes or even objects,
+Text::CSV already offers good functions to get lines out of a csv file, but in
+most cases you will want those lines to be parsed into hashes or even objects,
so this model just skips ahead and gives you objects.
Its basic assumptions are:
=item You do know what you expect to be in that csv file.
-This means first and foremost you have knowledge about encoding, number and
+This means first and foremost that you have knowledge about encoding, number and
date format, csv parameters such as quoting and separation characters. You also
know what content will be in that csv and what L<Rose::DB> is responsible for
it. You provide valid header columns and their mapping to the objects.
This module can handle multiplexed data of different class types. In that case
multiple profiles with classes and row identifiers must be given. Multiple
headers may also be given or read from csv data. Data must contain the row
-identifier in the first column and it's field name must be 'datatype'.
+identifier in the column named 'datatype'.
=back
Parse the data into objects and return those.
-This method will return list or arrayref depending on context.
+This method will return an arrayref of all objects.
=item C<get_data>
If not given, headers are taken from the first n lines of data, where n is the
number of different class types.
-In case of multiplexed data the first column must be named 'datatype'. This
-name must be given in the header.
+In case of multiplexed data there must be a column named 'datatype'. This
+column must be given in each header and must be at the same position in each
+header.
Examples:
classic data of one type:
[ 'name', 'street', 'zipcode', 'city' ]
- multiplexed data with two different types
+ multiplexed data with two different types:
[ [ 'datatype', 'ordernumber', 'customer', 'transdate' ],
[ 'datatype', 'partnumber', 'qty', 'sellprice' ] ]
-=item C<profile> [{profile => \%ACCESSORS, class => class, row_ident => ri},]
+=item C<profile> PROFILE_DATA
-This is an ARRAYREF to HASHREFs which may contain the keys C<profile>, C<class>
-and C<row_ident>.
+The profile mapping csv to the objects.
-The C<profile> is a HASHREF which may be used to map header fields to custom
+See section L</PROFILE> for information on this topic.
+
+=item C<ignore_unknown_columns>
+
+If set, the import will ignore unknown header columns. Useful for lazy imports,
+but deactivated by default.
+
+=item C<case_insensitive_header>
+
+If set, header columns will be matched against profile entries case
+insensitive, and on match the profile name will be taken.
+
+Only works if a profile is given, will die otherwise.
+
+If both C<case_insensitive_header> and C<strict_profile> is set, matched header
+columns will be accepted.
+
+=item C<strict_profile>
+
+If set, all columns to be parsed must be specified in C<profile>. Every header
+field not listed there will be treated like an unknown column.
+
+If both C<case_insensitive_header> and C<strict_profile> is set, matched header
+columns will be accepted.
+
+=back
+
+=head1 PROFILE
+
+The profile is needed for mapping csv data to the accessors in the data object.
+
+The basic structure is:
+
+ PROFILE := [ CLASS_PROFILE, CLASS_PROFILE* ]
+ CLASS_PROFILE := {
+ profile => { ACCESSORS+ },
+ class => $classname,
+ row_ident => $row_ident,
+ mapping => { MAPPINGS* },
+ }
+ ACCESSORS := $field => $accessor
+ MAPPINGS := $alias => $field
+
+The C<ACCESSORS> may be used to map header fields to custom
accessors. Example:
- [ {profile => { listprice => listprice_as_number }} ]
+ profile => {
+ listprice => 'listprice_as_number',
+ }
-In this case C<listprice_as_number> will be used to read in values from the
+In this case C<listprice_as_number> will be used to store the values from the
C<listprice> column.
-In case of a One-To-One relationsship these can also be set over
-relationsships by sparating the steps with a dot (C<.>). This will work:
+In case of a One-To-One relationship these can also be set over
+relationships by separating the steps with a dot (C<.>). This will work:
- [ {profile => { customer => 'customer.name' }} ]
+ customer => 'customer.name',
And will result in something like this:
$obj->customer($obj->meta->relationship('customer')->class->new);
$obj->customer->name($csv_line->{customer})
-But beware, this will not try to look up anything in the database. You will
+Beware, this will not try to look up anything in the database! You will
simply receive objects that represent what the profile defined. If some of
-these information are unique, and should be connected to preexisting data, you
+these information are unique, or should be connected to preexisting data, you
will have to do that for yourself. Since you provided the profile, it is
assumed you know what to do in this case.
If no profile is given, any header field found will be taken as is.
If the path in a profile entry is empty, the field will be subjected to
-C<strict_profile> and C<case_insensitive_header> checking, will be parsed into
-C<get_data>, but will not be attempted to be dispatched into objects.
+C<strict_profile> and C<case_insensitive_header> checking and will be parsed
+into C<get_data>, but will not be attempted to be dispatched into objects.
-If C<class> is present, the line will be handed to the new sub of this class,
-and the return value used instead of the line itself.
+C<class> must be present. A new instance will be created for each line before
+dispatching into it.
-C<row_ident> is a string to recognize the right profile and class for each data
-line in multiplexed data. It must match the value in the column 'dataype' for
-each class.
+C<row_ident> is used to determine the correct profile in multiplexed data and
+must be given there. It's not used in non-multiplexed data.
-In case of multiplexed data, C<class> and C<row_ident> must be given.
-Example:
- [ {
- class => 'SL::DB::Order',
- row_ident => 'O'
- },
- {
- class => 'SL::DB::OrderItem',
- row_ident => 'I',
- profile => {sellprice => sellprice_as_number}
- } ]
-
-=item C<ignore_unknown_columns>
+If C<mappings> is present, it must contain a hashref that maps strings to known
+fields. This can be used to add custom profiles for known sources, that don't
+comply with the expected header identities.
-If set, the import will ignore unkown header columns. Useful for lazy imports,
-but deactivated by default.
+Without strict profiles, mappings can also directly map header fields that
+should end up in the same accessor.
-=item C<case_insensitive_header>
-
-If set, header columns will be matched against profile entries case
-insensitive, and on match the profile name will be taken.
+With case insensitive headings, mappings will also modify the headers, to fit
+the expected profile.
-Only works if a profile is given, will die otherwise.
+Mappings can be identical to known fields and will be prefered during lookup,
+but will not replace the field, meaning that:
-If both C<case_insensitive_header> and C<strict_profile> is set, matched header
-columns will be accepted.
+ profile => {
+ name => 'name',
+ description => 'description',
+ }
+ mapping => {
+ name => 'description',
+ shortname => 'name',
+ }
-=item C<strict_profile>
+will work as expected, and shortname will not end up in description. This also
+works with the case insensitive option. Note however that the case insensitive
+option will not enable true unicode collating.
-If set, all columns to be parsed must be specified in C<profile>. Every header
-field not listed there will be treated like an unknown column.
-If both C<case_insensitive_header> and C<strict_profile> is set, matched header
-columns will be accepted.
+Here's a full example:
-=back
+ [
+ {
+ class => 'SL::DB::Order',
+ row_ident => 'O'
+ },
+ {
+ class => 'SL::DB::OrderItem',
+ row_ident => 'I',
+ profile => { sellprice => 'sellprice_as_number' },
+ mapping => { 'Verkaufspreis' => 'sellprice' }
+ },
+ ]
=head1 ERROR HANDLING
Note that the last entry can be off, but will give an estimate.
+Error handling is also known to break on new Perl versions and need to be
+adjusted from time to time due to changes in Text::CSV_XS.
+
=head1 CAVEATS
=over 4
projects / kivitendo-erp.git / blobdiff