We rely on shell history in our terminal operation. We search from our shell history and execute commands dozens times in a day.
However, shell history sometimes contains authorization tokens that we don't care while searching the commands. Some incremental fuzzy searchers have troubles when there are many random tokens in the shell history. Yeah, I know that I should not type a authorization token directly in the command line, but it's much easier than creating some shell script snippets.
Another hint to implement fillin is that programmers execute same commands switching servers.
We do not just login with ssh {{hostname}}, we also connect to the database with psql -h {{psql:hostname}} -U {{psql:username}} -d {{psql:dbname}} and to Redis server with redis-cli -h {{redis:hostname}} -p {{redis:port}}.
We switch the host argument from the localhost (you may omit this), staging and production servers.
Some command line tools allow us to login cloud services and retrieve data from our terminal.
Most of such cli tools accept an option to switching between the account.
For example, AWS command line tool has --profile option.
Other typical names of options are --region, --conf and --account.
When we specify these options directly, there are quadratic number of commands; the number of accounts times the number of actions.
The fillin allows us to save the command something like aws --profile {{aws:profile}} ec2 describe-instances so we'll not be bothered by the quadratic combinations of commands while searching through the shell history.
The core concept of fillin lies in separating the action (do what) in the command and the environment (to where).
With this fillin command line tool, you can
- make your commands reusable and it will make incremental shell history searching easy.
- fill in the template variables interactively and their history will be stored locally.
- invoke the same action switching multiple environment (local, staging and production servers, configuration path, cloud service accounts or whatever)
$ brew install itchyny/fillin/fillinReleases・itchyny/fillin - GitHub
$ go get -u github.com/itchyny/fillinThe interface of the fillin command is very simple.
Prepend fillin to the command and create template variables with {{...}}.
So the hello world for the fillin command is as follows.
$ fillin echo {{message}}
message: Hello, world! # you type here
Hello, world! # fillin executes: echo 'Hello, world!'The value of message variable is stored locally.
You can use the recently used value with the upwards key.
Note that in fish shell you can use square brackets like fillin echo [[message]].
The {{message}} (or [[message]] in fish shell) is a template part of the command.
As the identifier, you can use alphabets, numbers, underscore and hyphen.
Thus {{sample-id}}, {{SAMPLE_ID}}, {{X01}} and {{FOO_example-identifier0123}} are all valid template parts.
One of the important features of fillin is variable scope grouping.
Let's look into more practical example.
When you connect to PostgreSQL server, you can use:
$ fillin psql -h {{psql:hostname}} -U {{psql:username}} -d {{psql:dbname}}
[psql] hostname: localhost
[psql] username: example-user
[psql] dbname: example-dbWhat's the benefit of psql: prefix?
You'll notice the answer when you execute the command again:
$ fillin psql -h {{psql:hostname}} -U {{psql:username}} -d {{psql:dbname}}
[psql] hostname, username, dbname: localhost, example-user, example-db # you can select the most recently used entry with the upwards keyThe identifiers with the same scope name (psql scope here) can be selected as pairs.
You can input individual values to create a new pair after skipping the multi input prompt.
$ fillin psql -h {{psql:hostname}} -U {{psql:username}} -d {{psql:dbname}}
[psql] hostname, username, dbname: # just type enter to input values for each identifiers
[psql] hostname: example.org
[psql] username: example-org-user
[psql] dbname: example-org-dbThe scope grouping behaviour is useful with some authorization keys.
$ fillin curl {{example-api:base-url}}/api/1/example/info -H 'Authorization: Bearer {{example-api:access-token}}'
[example-api] base-url, access-token: example.com, accesstokenabcde012345The base-url and access-token are stored as tuples so you can easily switch between local, staging and production environment authorization.
Without the grouping behaviour, variable history searching will lead you to an unmatched pair of base-url and access-token.
Since the curl endpoint are stored in the shell history and authorization keys are stored in fillin history, we'll not be bothered by the quadratic number of the command history.
In order to have the benefit of this grouping behaviour, it's strongly recommended to prepend the scope name.
The psql: prefix on connecting to PostgreSQL database server, redis: prefix for Redis server are useful best practice in my opinion.
The terminal interface of fillin has problem with some shell functionalities.
For example, the following command gets stuck the terminal interface.
$ fillin echo {{message}} | jq .
{}^M^M^CThis is because the interface of fillin is rely on the standard output.
Instead of connecting the standard output of fillin to another command, pass the pipe character as an argument.
$ fillin echo {{message}} \| jq .
$ # or
$ fillin echo {{message}} '|' jq .Another solution is to quote the whole command.
$ fillin 'echo {{message}} | jq .'This is much simple and general method because you can use on subshells as well.
$ fillin 'echo $(cat {{foo}} {{bar}})'
$ fillin '(echo {{foo}}; cat {{bar}}) | grep func'The same problem occurs with redirection so escape > or quote the whole command.
$ fillin echo {{message}} \> /tmp/message
$ fillin echo {{message}} '>' /tmp/message
$ fillin 'echo {{message}} > /tmp/message'This command line tool is in its early developing stage. The user interface may be changed without any announcement.
This tool is not an encryption tool. The command saves the inputted values in a JSON file with no encryption.
Report bug at Issues・itchyny/fillin - GitHub.
itchyny (https://github.com/itchyny)
This software is released under the MIT License, see LICENSE.