App::combinesheets - command-line tool merging CSV and TSV spreadsheets


version 0.2.14


   combinesheets -h
   combinesheets -help
   combinesheets -man
   combinesheets -version

   combinesheets -config <config-file> -inputs <input-files> [<options>] [-outfile <output-file>]

      where <input-files> has the form: <input-ID>=<filename> [<input-ID>=<filename>...]
      where <options> are: -check


combinesheets is a command-line tool merging together two or more spreadsheets. The spreadsheets can be COMMA-separated or TAB-separated files, each of them having the first line with column headers. Data in one of the column (it can be a different column in each input spreadsheet) serve to match lines. For example, having two spreadsheets, PERSON and CAR, with the following contents:


   Surname      First name  Sex  Age  Nickname
   Novak        Jan         M    52   Honza
   Gudernova    Jitka       F    56
   Senger       Martin      M    61   Tulak


   Model  Year  Owned by
   Praga  1936  Someone else
   Mini   1968  Gudernova
   Skoda  2002  Senger

we want to merge these spreadsheet by Surname in persons.tsv and by Owned by in cars.tsv. There are two possible results, depending which spreadsheet is used as the first one (a primary one). If the persons.tsv is the first, the result will be (which columns are included in the result will be described later in this document):

   combinesheets -cfg config.cfg -in PERSON=persons.tsv CAR=cars.csv

   First name  Surname    Model  Sex  Nickname  Age  Year  Owned by
   Jitka       Gudernova  Mini   F              56   1968  Gudernova
   Jan         Novak             M    Honza     52
   Martin      Senger     Skoda  M    Tulak     61   2002  Senger

Or, if the cars.tsv is the first, the result will be:

   combinesheets -cfg config.cfg -in CAR=cars.csv PERSON=persons.tsv

   First name  Surname    Model  Sex  Nickname  Age  Year  Owned by
   Jitka       Gudernova  Mini   F              56   1968  Gudernova
   Martin      Senger     Skoda  M    Tulak     61   2002  Senger
                          Praga                      1936  Someone else

Of course, if both input spreadsheets have only the matching lines, both results will be the same (it will not matter which one of them is considered the primary one).

The rows in the resulting spreadsheet are sorted by values in the column that was used as a matching column in the primary input.

The information which columns should be used to match the input spreadsheets and which columns should appear in the resulting spreadsheet is read from a configuration file (see the -config - or -cfg - argument).

The command-line arguments and options can be specified with single or double dash. Most of them can be abbreviated to the nearest non-biased length. They are case-sensitive.

Duplicated values in the matching columns

If there are repeated (the same) values in the column that serves as matching criterion then the resulting spreadsheet will have as many output lines (for a particular matching value) as is the number of all combinations of the lines with that matching values in all input spreadsheets. For example, let's have books.tsv and authors.tsv, assuming that a book can have more authors and any author can contribute to any number of books:

   Title   Note    Author
   Book 1  from B1-a       Kim
   Book 2  from B2-b       Kim
   Book 3  from B3-c       Katrin
   Book 1  from B1-d       Blanka
   Book 2  from B2-e       Katrin

   Age     Name
   28      Kim
   20      Katrin
   30      Blanka
   50      Lazy author

The output (again, depending on which input is considered a primary input) will be (a list of included column is defined in the configuration file - see later):

   combinesheets -cfg books_to_authors.cfg -in BOOK=books.tsv AUTHOR=authors.tsv

   Name    Title   Age Note
   Blanka  Book 1  30  from B1-d
   Katrin  Book 3  20  from B3-c
   Katrin  Book 2  20  from B2-e
   Kim     Book 1  28  from B1-a
   Kim     Book 2  28  from B2-b

   combinesheets -cfg books_to_authors.cfg -in AUTHOR=authors.tsv BOOK=books.tsv

   Name        Title   Age  Note
   Blanka      Book 1  30   from B1-d
   Katrin      Book 3  20   from B3-c
   Katrin      Book 2  20   from B2-e
   Kim         Book 1  28   from B1-a
   Kim         Book 2  28   from B2-b
   Lazy author         50


Additionally to the merging columns from one or more spreadsheets, this script can also add completely new columns to the resulting spreadsheet, the columns that do not exist in any of the input spreadsheet. Such columns are called calculated columns.

Each calculated column is created either by an external, command-line driven, program, or by a Perl subroutine. In both cases, the user must create (write) such external program or such Perl subroutine. Therefore, this usage is meant more for developers than for the end users.

Note that this advanced feature is meant only for new columns, not for new rows. Therefore, it cannot be used, for example, to create rows with totals of columns.

Calculated columns by external programs

If specified, an external program is invoked for each row. It can be specified either by a keyword PROG or by a keyword PROGS - see syntax in the configuration section. In both cases, the value of the standard output of these programs become the value of the calculated column (a trailing newline of this standard output is removed and other newlines are replaced by spaces).

A program defined by the PROGS is called without any arguments (S in PROGS stands for a Simple). That's why it does not have any knowledge for which row it has been invoked. Its usage is, therefore, for column values that are not dependent on other values from the spreadsheet. For example, for the cars.tsv shown above, you can add a column Last updated by calling a UNIX program date - again, see an example the configuration section.

A program defined by the PROG is called with one argument which is a filename. This file contains the current row; each of its lines has two, TAB-separated, fields. The first field is the column name and the second field is the column value. For example, when processing the last row of the cars.tsv given above, the file will have the following content:

   Model       Skoda
   Year        2002
   Owned by    Senger

The files are only temporary and will be removed when combinesheets finishes.

Calculated columns by a Perl subroutine

If specified by the keyword PERL, a Perl subroutine is called for each row with the three arguments:

  1. A hashref with information about the current column. Not often used but may be handy if the same subroutine deals with more columns and, therefore, needs to know for which column it was invoked. See the flights example in the configuration section.

  2. An arrayref with all column names.

  3. An arrayref with all column values - in the same order as the column names.

Actually, depending how the subroutine is defined in the configuration, it may get as the first argument the module/class name where it belongs to. If you define it like this:

   PERL   Module::Example::test

the test subroutine is called, indeed, with the three arguments as described above. However, if your definition is rather:

   PERL   Module::Example->test

then the test subroutine is considered a Perl method and its first argument is the module/class name. It is up to you to decide how you want/need to write your functions. Again, an example is available in the configuration section.

The return value of the subroutine will become a new value in the calculated column. Do not return undef but rather an empty string if the value cannot be created.

What is an advantage of writing my own module/package if I can simply write an external program (perhaps also in Perl) doing exactly the same? The Perl module stays in the memory for the whole time of processing all input rows and, therefore, you can re-use some calculations done for the previous rows. An example about it (flights) is given in the configuration section.


-config <config-file>

A filename with a configuration file. This is a mandatory parameter. The configuration file describes:

  • which columns in individual input spreadsheets should be included in the resulting spreadsheet,

  • what names should be given to the resulting columns

  • in which order should be the columns in the resulting spreadsheet

  • which columns should be used to match individual lines,

The configuration file is a TAB-separated file (with no header line). Empty lines and lines starting with a "#" character are ignored. Each line has two columns, in some cases there is an optional third column. Here is a configuration file used in the example above:

   # Columns to match records from individual inputs
   MATCH   PERSON=Surname
   MATCH   CAR=Owned by
   MATCH   CHILD=Parent

   # Columns - how they be in rows
   PERSON  First name
   PERSON  Surname
   CAR     Model
   PERSON  Sex
   CHILD   Name
   CHILD   Born
   PERSON  Nickname
   PERSON  Age
   CAR     Year
   CAR     Owned by

The first column is either a reserved word MATCH, or an identifier of an input spreadsheet. There are also few other reserved words - see more about them a bit later.

The identifier can be almost anything (and it does not appear in the input spreadsheet itself). It is also used in the command-line argument -inputs where it corresponds to a real file name of the input. The lines with identifiers define what columns will be in the result: the second column is the header of the wanted columns and an optional third column (not used in the example above) is the header used in the result. The resulting columns will be in the same order as are these lines in the configuration file.

The reserved word MATCH is used to define how to match lines in the input spreadsheets. The format of its second column is:


There should be one MATCH line for each input spreadsheet. The data in the column defined by the "column-header" will be used to find the corresponding lines. In our example, the data in the column Surname in the persons.tsv will be matched with the data in the column Owned by in the cars.tsv (the rows having the same values in these two columns will be merged into one resulting row).

Advanced configuration

If you want to add so-called calculated columns as described in the "ADVANCED USAGE" you need to use few additional reserved words in the configuration file. These words are PROG, PROGS and/or PERL. They are used in the place where the new calculated column should be placed. Their lines have the program name or the Perl subroutine name in the second column, and they have mandatory third column with the resulting name of the calculated column.

For example, we wish to add two columns to the input spreadsheet cars.tsv. The input file (the same as in the introduction) is:

   Model  Year  Owned by
   Praga  1936  Someone else
   Mini   1968  Gudernova
   Skoda  2002  Senger

We wish to add a column Car age that shows the difference between the actual year and the value from the Year column. We have a shell script doing it:

   YEAR=`grep Year $1 | cut -f2`
   NOW=`date +%Y`
   echo $(($NOW-$YEAR))

The configuration file cars.cfg (assuming that we want the other columns to remain the same) is:

   MATCH   CAR=Owned by

   CAR     Owned by
   CAR     Model
   CAR     Year
   PROG  Car age

When we run:

   combinesheets -config cars.cfg -inputs CAR=cars.tsv

we get this result:

   Owned by        Model   Year    Car age
   Gudernova       Mini    1968    44
   Senger          Skoda   2002    10
   Someone else    Praga   1936    76

You can see that there is no need to use combinesheets for really combining more sheets, an input can be just one sheet.

Another example adds a fixed column to the same input, a column named Last updated that gets its value from a UNIX command date. This program does not get any information which row it has been invoked for. The configuration file is now (note the new line with the PROGS):

   MATCH   CAR=Owned by

   CAR     Owned by
   CAR     Model
   CAR     Year
   PROG  Car age
   PROGS   date    Last updated

and the result is now:

   Owned by        Model   Year    Car age   Last updated
   Gudernova       Mini    1968    44        Mon Feb 27 12:32:04 AST 2012
   Senger          Skoda   2002    10        Mon Feb 27 12:32:04 AST 2012
   Someone else    Praga   1936    76        Mon Feb 27 12:32:04 AST 2012

The last possibility is to call a Perl subroutine - using the reserved word PERL in the configuration file. Let's have an input spreadsheet (flights.tsv) with data about flights:

   Date         Flight    Airport From      Airport To
   2009-01-18   AY838     London LHR        Helsinki Vantaa
   2009-01-22   AY839     Helsinki Vantaa   London LHR
   2009-03-15   NW2       Manila            Tokyo Narita
   2009-03-21   NW1       Tokyo Narita      Manila
   2011-05-06   SV326     Sharm El Sheik    Jeddah
   2011-07-31   RJ700     Amman             Jeddah
   2011-09-21   ME369     Jeddah            Beirut
   2011-09-24   ME368     Beirut            Jeddah
   2011-12-02   EZY3064   Prague            London Stansted
   2011-12-09   EZY3067   London Stansted   Prague
   2012-01-26   MS663     Cairo             Jeddah

We want to add columns with the international airport codes for both Airport From and Airport To. The new columns will be named Code From and Code To. The Perl subroutine will use a web service to find the code. The subroutine will use a closure that will remember already fetched codes so the web service does not need to be called several times for the same airport name.

The configuration file flights.cfg is:

   MATCH   FLY=Date

   FLY     Date
   FLY     Flight
   FLY     Airport From
   PERL    Airport->find_code      Code From
   FLY     Airport To
   PERL    Airport->find_code      Code To

The name of the subroutine is attached to the module where it comes from by either :: or -> notation.

The invocation is:

   combinesheets -config flights.cfg -inputs FLY=flights.tsv

The full code for the module Airport, the file is here:

   package Airport;
   use warnings;
   use strict;

   use LWP::Simple;
   use JSON;

   # preparing a closure in order not to fetch the same airport code again and again
   my $already_found = make_already_found();
   sub make_already_found {
      my $already_found = {};
      return sub {
         my ($airport_name, $airport_code) = @_;
         if (exists $already_found->{$airport_name}) {
            if ($airport_code) {
                $already_found->{$airport_name} = $airport_code;
            return $already_found->{$airport_name};
         } else {
            $already_found->{$airport_name} = ($airport_code ? $airport_code : 1);
            return 0;

   sub find_code {
      my ($class, $column, $header_line, $data_line) = @_;

      my $column_with_airport_name = $column->{ocol};
      $column_with_airport_name =~ s{Code}{Airport};

      my $airport_name;
      for (my $i = 0; $i < @$header_line; $i++) {
         if ($header_line->[$i] eq $column_with_airport_name) {
            $airport_name = $data_line->[$i];
      return '' unless $airport_name;

      # now we have an airport name...
      my $airport_code = $already_found->($airport_name);
      return $airport_code if $airport_code;

      #... go and find its airport code
      $airport_code = '';
      my $escaped_airport_name = $airport_name;
      $escaped_airport_name =~ tr{ }{+};
      my $url = "$escaped_airport_name&fmt=json";
      my $content = get ($url);
      warn "Cannot get a response for '$url'\n"
         unless defined $content;
      my $json = JSON->new->allow_nonref;
      my $data = $json->decode ($content);
      foreach my $code (@$data) {
         $airport_code .= $code->{code} . ",";
      chop ($airport_code) if $airport_code;  # removing the trailing comma

      $already_found->($airport_name, $airport_code);
      return $airport_code;

When run it creates the following output. Note that some airports have more than one code because the name was ambiguous. Well, this is just an example, isn't it?

   Date         Flight    Airport From      Code From   Airport To       Code To
   2009-01-18   AY838     London LHR        LHR         Helsinki Vantaa  HEL
   2009-01-22   AY839     Helsinki Vantaa   HEL         London LHR       LHR
   2009-03-15   NW2       Manila            MXA,MNL     Tokyo Narita     NRT
   2009-03-21   NW1       Tokyo Narita      NRT         Manila           MXA,MNL
   2011-05-06   SV326     Sharm El Sheik    SSH         Jeddah           JED
   2011-07-31   RJ700     Amman             ADJ,AMM     Jeddah           JED
   2011-09-21   ME369     Jeddah            JED         Beirut           BEY
   2011-09-24   ME368     Beirut            BEY         Jeddah           JED
   2011-12-02   EZY3064   Prague            PRG         London Stansted  STN
   2011-12-09   EZY3067   London Stansted   STN         Prague           PRG
   2012-01-26   MS663     Cairo             CAI,CIR     Jeddah           JED
-inputs <input_ID=<filename> [<input_ID>=<filename>...]>

Each -inputs can have one or more file names, and there can be one or more -inputs arguments. It defines what are the input spreadsheets and how they are identified in the configuration file (see the -config argument). For example, the inputs for our example above can be specified in any of these ways:

   -inputs PERSON=persons.tsv -inputs CAR=cars.tsv
   -inputs PERSON=persons.tsv CAR=cars.tsv
   -inputs PERSON=persons.tsv,CAR=cars.tsv

The first file name is considered to be the primary input (see the description above): the resulting spreadsheet will have the same number of lines as the primary input.

The file names ending with the .csv are considered to be in the COMMA-separated formats, all others are considered to be TAB-separated.

This is a mandatory parameter.

-outfile <output-file>

An optional parameter specifying a filename of the combined result. By default, it is created on STDOUT. It is always in the TAB-separated format.


This option causes that the configuration file and the input files (only their header lines will be read) will be checked for errors but no resulting spreadsheet will be created.


Not yet implemented.

General options

Print a brief usage message and exits.


Print a brief usage message with options and exit.


Print a full usage message and exit.


Print the version and exit.



It contains a path that is used when looking for external programs (when the reserved words PROG or PROGS are used). For example, the examples directory in the source distribution of this package has an external program The full invocation can be done by:

   COMBINE_SHEETS_EXT_PATH=examples bin/combinesheets -cfg examples/cars.cfg --inputs CAR=examples/cars.csv


In order to run this tool you need Perl and the following Perl modules to be installed:


Optionally (if your configuration file uses the reserved word PROG or PROGS for calculated columns):



  • Columns are identified by their header names. There is no way to identify them simply by their order (column number).

  • The input spreadsheet are read first into memory. Which may be a problem with really huge spreadsheets.

  • The inputs can be COMMA-separated or TAB-separated. It would be perhaps nice to allow also the Excel spreadsheets.

  • Comparing header names and rows is case-sensitive only. There is a plan to implement the option -ignorecases,

Some of these missing features may be implemented later.


You can find documentation for this module with the perldoc command.

    perldoc App::combinesheets

You can also look for information at:


Martin Senger <>


This software is copyright (c) 2013 by Martin Senger, CBRC - KAUST (Computational Biology Research Center - King Abdullah University of Science and Technology) All Rights Reserved..

This is free software; you can redistribute it and/or modify it under the same terms as the Perl 5 programming language system itself.