App::RecordStream - recs - A system for command-line analysis of data


A set of programs for creating, manipulating, and outputting a stream of Records, or JSON hashes. Inspired by Monad.


Quick, standalone bundle

The quickest way to start using recs is via the minimal, standalone bundle:

  curl -fsSL > recs
  chmod +x recs
  ./recs --help

This is also known as the "fatpacked" recs.


You can also install recs from CPAN as App::RecordStream:

  cpanm --interactive App::RecordStream

Using cpanm in interactive mode will prompt you for optional feature support. Other CPAN clients such as cpan and cpanp also work fine, but you can't opt to use any optional features (just like cpanm in non-interactive mode). A kitchen-sink install of App::RecordStream looks like:

  cpanm --with-recommends --with-all-features App::RecordStream

If you don't have cpanm itself, you can install it easily with:

  curl -fsSL | perl - App::cpanminus


The recs system consists of three basic sets of commands:

  • Input commands responsible for generating streams of record objects

  • Manipulation commands responsible for analyzing, selecting, and manipulating records

  • Output commands responsible for taking record streams and producing output for humans or other programs

These commands can interface with other systems to retrieve data, parse existing files, or just regex out some values from a text stream.

Commands are run using recs command [options and arguments]. If you're using a CPAN-based install, you may also run commands directly as recs-command, though this is no longer recommended for forwards compatibility. Both installation methods provide a top-level recs executable which dispatches to commands, so this is the preferred invocation style.

The core recs commands are briefly summarized below, and you can list all available commands by running recs --list.

To read more about each command, run recs command --help. Longer documentation is available as recs command --help-all or perldoc recs-command. For example, to read more about "fromcsv", you might run any of the following:

  recs fromcsv --help
  recs fromcsv --help-all
  perldoc recs-fromcsv


Input Generation


Produces records from a csv file/stream


Produces records for a db table, or from a SELECT statement into a db.


Matches input streams against a regex, puts capture groups into hashes


Generate a record stream from a MongoDB query.


Matches input streams against several regexes, puts capture groups into the record


Splits input stream on a delimiter


Generate records from the process tree


Produces records for an optionally paginated atom feed.


Produces records for an XML document.


Produces records from input streams containing loosely formed key/value pairs


Produces records from packet capture files (.pcap) as made by tcpdump

Stream Manipulation


Annotate records with common fields, will memoize additions to speed up common annotations


Perforce aggregation operations on records. Group by a field, get an average, sum, correlation, etc. Very powerful


Transform values into deltas between adjacent records


Eval a string of Perl against each record


Flatten records of input to one level


Select records for which a string of Perl evaluates to true.


Take records, grouped by keys, and run a separate recs command for each group.


Based on a time field, tag records with a normalized time, i.e. every 5 minute buckets


Perform an inner join of two record streams. Associate records in one stream with another stream.


Filter to a range of matching records with paired Perl snippets --start and --end.


Sort records based on keys, may specify multiple levels of sorting, as well as numerical or lexical sort ordering


Outputs the top n records. You may segment the input based on a list of keys such that unique values of keys are treated as distinct input streams. This enables top n listings per value groupings.


Perform a block of Perl on each record, which may modify the record, Record is then output


Perform a block of Perl on each record to generate a record stream, which is then output with a chain link back to the original record.

Output Generation


Inserts records into a DBI supported SQL database. Will create a local SQLite database by default


Generates correctly quoted CSV files from record streams.


Create a graph of field values in a record using GNU Plot.


Pretty prints a table of results.


Prints out an HTML table of the record stream


Prettily prints records, one key to a line, great for making sense of very large records


Prints a multi-dimensional (pivot) table of values. Very powerful.


Many of the commands above take key arguments to specify or assign to a key in a record. Almost all of the places where you can specify a key (which normally means a first level key in the record), you can instead specify a key spec.

A key spec may be nested, and may index into arrays. Use a / to nest into a hash and a #NUM to index into an array (i.e. #2)

An example is in order, take a record like this:

  {"biz":["a","b","c"],"foo":{"bar 1":1},"zap":"blah1"}
  {"biz":["a","b","c"],"foo":{"bar 1":2},"zap":"blah2"}
  {"biz":["a","b","c"],"foo":{"bar 1":3},"zap":"blah3"}

In this case a key spec of foo/bar 1 would have the values 1, 2, and 3 respectively.

Similarly, biz/#0 would have the value of a for all 3 records

Fuzzy matching

You can also prefix key specs with @ to engage the fuzzy matching logic. Matching is tried like this, in order, with the first key to match winning:

1. Exact match (eq)
2. Prefix match (m/^/)
3. Match anywhere in the key (m//)

Given the above example data and the fuzzy key spec @b/#2, the b portion would expand to biz and 2 would be the index into the array, so all records would have the value of c.

Simiarly, @f/b would have values 1, 2, and 3.


The data stream format of the recs commands is JSON hashes separated by new lines. If you wish to write your own recs command in your own language, just get a JSON parser and you should be good to go. The recs commands use JSON::MaybeXS.

If you name your command as recs-mycommand and put it somewhere in your PATH environment variable, the recs command will dispatch to it when called as recs mycommand. It will also be included in recs --list output.

If you want to write your new command in Perl, you can use the same Perl API that the standard recs toolkit uses. See the various App::RecordStream::Operation subclasses. Once your new operation class is installed in perl's library paths, recs will find it automatically without the need for any executable command shim.


  # look in the custom access log for all accesses with greater than 5 seconds,
  # display in a table
  cat access.log \
    | recs fromre --fields ip,time '^(\d+).*TIME: (\d+)' \
    | recs grep '$r->{time} > 5' \
    | recs totable


Each of the commands discussed have a --help mode available to print out usage and examples for the particular command. See that documentation for detailed information on the operation of each of the commands. Also see some other man pages:


Benjamin Bernard <>

Keith Amling <>

Thomas Sibley <>


Copyright 2007–2017 by the AUTHORS

This software is released under the MIT and Artistic 1.0 licenses.