A CLI tool for Kafka

Plain text output is often very useful, and allows for output to be piped to standard unix tools for manipulation. However, the current "plain" output format has some decoration that impedes this, e.g.

$ trubka list topics -f plain
Topics (96)
───────────
 ● __consumer_offsets
 ● topic1
 ● topic2
 ● topic3
[...]

 Total
─────
    Topics: 96
Partitions: 180

Please change plain, or add a different output format, that is just a simple plain-text list, with no decoration, headers, or footers, e.g.

$ trubka list topics -f plain
topic1
topic2
topic3

Consider adding JSON output format. For better or worse it's now the standard data interchange, and with tools like jq is easily manipulated, filtered, etc. This would be especially useful for creating input to tools like kafka-reassign-partitions.sh. For reference here's what that looks like, first for input to --topics-to-move-json-file:

{
  "topics": [
    {
      "topic": "topic1" 
    },
    {
      "topic": "topic2" 
    },
    {
      "topic": "topic3" 
    }
  ],
  "version": 1
}

Then for input to --reassignment-json-file:

{
  "version": 1,
  "partitions": [
    {
      "topic": "topic1",
      "partition": 0,
      "replicas": [1, 2],
      "log_dirs": ["any","any"]
    },
    {
      "topic": "topic2",
      "partition": 0,
      "replicas": [2, 3],
      "log_dirs": ["any","any"]
    },
    {
      "topic": "topic3",
      "partition": 0,
      "replicas": [1, 3],
      "log_dirs": ["any","any"]
    }
  ]
}

Is your feature request related to a problem? Please describe. It would be nice to have a --to parameter in the consumer. One benefit is that you can pipe your consumed messages to a file and the consumer will finish once the --to condition is reached. Another benefit is that if you're searching through messages, it allows you to narrow your search.

Describe the solution you'd like The ability to stop consuming by timestamp, by latest offset at the time of starting the consumer, and/or by specific offsets by partition.

For example you could consume messages between 3pm and 4pm like:

$ trubka consume plain <topic name> --brokers <broker:port> --from 22-02-2020T15:00:00 --to 22-02-2020T16:00:00

Or you could consume messages from 3pm until now like:

$ trubka consume plain <topic name> --brokers <broker:port> --from 22-02-2020T15:0:00 --to newest

Describe the bug Setting TRUBKA_PROTO_ROOT environment variable still gives an error about requiring the --proto-root flag.

To Reproduce

$ echo $TRUBKA_PROTO_ROOT
/Users/jeff/dev/platypus

$ trubka consume proto --topic-filter wifi --proto-filter Blip --brokers localhost:9092 --kafka-version 2.2.0 -i
ERROR: Required flag --proto-root not provided

Expected behavior If I provide this environment variable I shouldn't have to provide --proto-root. According to the wiki:

It is possible to ask Trubka to read the cli flags from the system's environment variables. The flags must be in TRUBKA_FLAG_NAME format. For example the value of --proto-root parameter can be read from TRUBKA_PROTO_ROOT if it's provided.

I was trying to generate a boolean value, and I was getting an error, after checking the source code I could see what it was a bug. One missing quote is missing from the boolean regex, as a result the produced output still had the quote, producing an invalid JSON document

Today I needed to review the retentions configured for all our topics. Rather than iterating over all the topics with the shell I thought it might make sense to have the list command grab the metadata too. I added -c and -o options, same as describe, that causes it to grab the config and offsets and stick it into a Metadata field.

Right now it only shows up in the json output, which probably makes sense, since the table would be unwieldy with all the metadata (but it would be really nice to be able to choose columns of interest).

You may have some better idea or preference on how this is done, I'm open to ideas, but this met my need so I wanted to offer it back.

Example usage:

$ trubka list topics  -c -t ingest\.dcache\.billing -f json | \
jq -c '.[] | {"name":.name, "retention_days":(.metadata.configurations | from_entries | ."retention.ms" | tonumber /1000/3600/24)}'
{"name":"ingest.dcache.billing","retention_days":7}
{"name":"ingest.dcache.billing.cms-disk","retention_days":7}
{"name":"ingest.dcache.billing.cms-tape","retention_days":30}

Is your feature request related to a problem? Please describe. When using trubka consume proto if you have a protobuf with a field of type bytes, the field will be formatted as base64, but I would like to look at them in hex string format. There is an option to format your entire message as hex using -F hex but not just bytes fields.

Describe the solution you'd like An option to output in -F json-indent format, but with individual fields of type bytes as hex strings. For example adding a flag --bytes-format [base64|hex]. You could take this concept even further and have custom formatting for any type with some kind of configuration file. For example you may want timestamps in a different format as well.