log-defer-viz - command-line utility for rendering log messages created by Log::Defer


Log::Defer is a module that creates structured logs. The Log::Defer documentation explains structured logging and its benefits over ad-hoc logging.

This module installs a command-line script that parses structured logs created by Log::Defer and displays them in a readable manner.


The fastest way to install log-defer-viz is with cpanminus:

    curl -sL | sudo perl - Log::Defer::Viz



    $ cat file.log | log-defer-viz
    $ log-defer-viz < file.log
    $ log-defer-viz file.log
    $ log-defer-viz -F file.log # continuously tail file
    $ log-defer-viz file.log file2.log
    $ log-defer-viz archived.log.gz more_logs.bz2


    $ log-defer-viz --input-format=json  ## default is newline separated JSON
    $ log-defer-viz --input-format=sereal  ## Sereal::Decoder (not impl)
    $ log-defer-viz --input-format=messagepack  ## Data::MessagePack (not impl)
    $ log-defer-viz --input-format=storable  ## Storable (not impl)

Note: The only input format currently implemented is newline-separated JSON.


    $ log-defer-viz  ## by default shows error, warn, and info logs
    $ log-defer-viz -v  ## verbose mode (adds debug logs and more)
    $ log-defer-viz --debug  ## show debug logs
    $ log-defer-viz --quiet  ## only errors and warnings
    $ log-defer-viz --verbosity 25  ## numeric verbosity threshold
    $ log-defer-viz --nowarn  ## muffle warn logs (so show error and info)
    $ log-defer-viz --nologs  ## don't show log section
    $ log-defer-viz --nocolour  ## turn off terminal colours
    $ log-defer-viz --preserve-newlines # don't indent multi-line log messages


    $ log-defer-viz --timer-columns 80  ## width of timer chart
    $ log-defer-viz --since-now  ## show relative to now times
                                 ##   like "34 minutes ago"
    $ log-defer-viz --notimers  ## don't show timer chart
    $ log-defer-viz --tz UTC  ## show times in UTC, not local


Applications can optionally log information in a "data" hash. This information is mostly designed to be extracted by programs so log-defer-viz doesn't display it by default. Use the --data option to display it anyway, and the --data-format option to choose the format to display it in. The available formats are pretty-json, json, yaml, and dumper.

    $ log-defer-viz --data  ## show data section. default is pretty-json
    $ log-defer-viz --data-format=json  ## compact, not pretty
    $ log-defer-viz --data-format=dumper  ## Data::Dumper
    $ log-defer-viz --data-only  ## only show data


As described in detail in their respective sections below, --grep, --map, and --reduce allow flexible selection and manipulation of your log data using arbitrary perl code. In the provided perl code, $_ refers to the log entry as a hash-reference. In --reduce there is also a special $o output variable.

    $ log-defer-viz --grep '$_->{data}'  ## grep for records that have a data section.
                                         ## $_ is the entire Log::Defer entry.
    $ log-defer-viz --map '$_->{data}->{username}'  ## Extract username from data
    $ log-defer-viz --reduce '$o->{ $_->{data}->{ip_addr} }++'  ## Count IP addresses
    $ log-defer-viz --pass-through  ## After grepping, print a valid log-defer stream


The count parameter tallies values found in log file. The arguments can be keys in the data section or arbitrary perl code. Multiple values are accepted. Note: This feature is mostly obsoleted by the --reduce feature but is kept for backwards compatibility and because it can be quite convenient.

    $ log-defer-viz --data --count ip_address ## display how many log lines for each ip address
    $ log-defer-viz --data --count ip_address --count '$_->{data}->{login_info}->{username}'


    $ log-defer-viz --help  ## the text you are reading now
    $ log-defer-viz --sort-time  ## sort by start time


As shown above, there is a --grep command-line option. This lets you filter log messages using arbitrary perl code. If the expression returns true, the log message is processed and displayed as usual.

Being able to do this easily is an important advantage of structured logs. With unstructured logs it is often difficult to extract all of the information related to a request and nothing else.

For example, here is how to grep for all requests that took longer than 500 milliseconds:

    $ log-defer-viz --grep '$_->{end} > .5' server.log

Depending on your underlying storage format, it may be meaningful to grep before passing to log-defer-viz (usually for performance reasons). Currently the only supported storage format is newline-separated JSON which is designed to be pre-grepable. If your search string appears anywhere in the object, the entire log message will be displayed:

    $ grep app.log | log-defer-viz

The final and most error-prone way to grep Log::Defer logs is to grep the unstructured output of log-defer-viz (not recommended):

    $ log-defer-viz app.log | grep

In general, --grep can be combined with other switches that take expressions such as --map and --reduce. In these cases, the greping will occur first.


Similar to --grep, there is also a --map command-line option. If this option is passed in, the only output is whatever your --map expression returns.

For example, if you are putting the PID into the data section with $log->data->{pid} = $$, then you can extract the PID like so:

    $ log-defer-viz --map '$_->{data}->{pid}' < app.log

Join together fields with a pipe:

    $ log-defer-viz --map 'join "|", $_->{data}{pid}, $_->{start}' < app.log

Make dates readable (localtime in scalar context makes a timestamp readable):

    $ log-defer-viz --map 'join "|", $_->{data}{pid}, "".localtime($_->{start})' < app.log
    9765|Fri Mar  1 14:37:53 2013
    9768|Fri Mar  1 15:00:38 2013
    9771|Fri Mar  1 15:24:42 2013

As with --grep, you have access to any perl functions you might need. Also, you can combine --map and --grep. The grep filtering will be applied before the mapping.

For example, here is how to do a "pass-through grep" where the output is another valid JSON-encoded Log::Defer file:

    $ export USER=jimmy
    $ log-defer-viz -g "_->{data}{username} eq '$USER'" \
                    -m "encode_json _"                  \
                    < requests.log                      \
                    > jimmys-requests.log

Note that the above also demonstrates two shortcut features: First, the -g and -m switches are abbreviations for --grep and --map respectively. Second, in grep and map expressions the _ function is an abbreviation for $_. Although this is one character shorter, the main reason this exists is so that you can use double-quoted strings without having to worry about escaping $ characters from your shell.

Instead of using -m "encode_json $_", there is a --pass-through option that is more efficient since it doesn't pointlessly re-encode the log message.


Unlike --map which outputs the same number of values that were input, and --grep that can return any number up to that amount, --reduce returns exactly one value.

While the expression passed in as an argument to --reduce is being evaluated, there is a special variable called $o available. The initial value of this variable is an empty hash ({}) and the value stored there is preserved across all invocations of your reduce expression. This means it can be used for counters and accumulators and such.

The most straightforward use-case is the simple tally or count. For example, here is how you could mimic the (somewhat deprecated) --count feature and get the sum totals of the different HTTP status codes present in the log file:

    $ log-defer-viz --reduce '$o->{ $_->{data}->{status_code} }++' \
        "200": 10293,
        "404": 392,
        "304": 3012,
        "500": 2

Here is how to count how many log entries contain at least one error message (log level 10 or lower, see Log::Defer):

    --reduce '$o->{errors}++ if grep { $_->[1] <= 10 } @{ $_->{logs} }'

As with --data and --count, the default output format is prettified JSON but you can change that with the --data-format option.



Log::Defer::Viz github repo

The log-defer-viz is also useful for visualising logs created by Michael Pucyk's LogDefer Python module (since it outputs the same format)


Doug Hoyte, <>


Matt Phillips, <>

Mike P

Mike R


Thanks to the above and also to everyone else who has given feedback or suggestions.


Copyright 2013-2016 Doug Hoyte and contributors.

This module is licensed under the same terms as perl itself.