OPY command that makes Python easy on CLI

Sun Sep 8 17:41:20 JST 2019 (modified: Sun Sep 8 18:57:30 JST 2019)
views: 159, keywords: opy, Python, onelinar, shell-gei

  このエントリーをはてなブックマークに追加 
   

summary

Since I want to use Python instantly on CLI, I have created opy command.

design and implementation

works like a line-oriented language

AWK (awk as a command) has been frequently used for one liners because it is a line-oriented language. When text data is given to awk from the standard input, it splits each line to a list of fields without any option or explicit code. This is an example.

$ seq 3 | awk '{print $1*3}'  # triple the number of the field one and output
3
6
9

Each number is set to $1 automatically. This mechanism enables us to write a short one liner.

On the other hand, when we try to do the same thing with Python, we need a longer code.

$ seq 3 | python3 -c 'import sys;print("\n".join([ str(int(x)*3) for x in sys.stdin ]))'
3
6
9

It is clearly unsuitable for oneliner.

However, Python has the list data type. Then AWK also treats fields of each line as elements of a list. So I thought the action of AWK can be replaced to the list of Python.

I have implemented opy command, which works with this idea. This is an example.

$ seq 3 | opy '[F1*3]'
3
6
9

As an action, the code in the argument only gives the list. The command prints the list as awk does.

 Opy also accepts the pattern. Differently from AWK, we need : between a pattern and an action.

### opy ###
$ seq 3 | opy 'F1%2:[F1*3]'
3
9

I also show equivalent codes with AWK and Python3.

### AWK ###
$ seq 3 | awk '$1%2{print $1*3}'
3
9
### Python3 ###
$ seq 3 | python3 -c 'import sys;print("\n".join([ str(int(x)*3) for x in sys.stdin if int(x)%2 ]))'
3
9

When there are more than one elements in the list, the elements are joined with spaces.

$ seq 3 | opy 'F1%2:["%d:odd"%F1, F1*3]'
1:odd 3
3:odd 9

Opy also has a normal action mode. When we use it, we need to put sentences in curly brackets.

$ seq 3 | opy '{print(F1, end="")}'
123

grammer

Over the Python grammer, following rules are added. They enables us to write the AWK like action/pattern rules.

<rules> ::= <rule> | <rule> ";" <rules>
<rule> ::= <pattern> | <pattern> ":" <action> | <action>
<action> ::= <list action> | <normal action>
<pattern> ::= <boolean expression of Python> | "B" | "BEGIN" | "E" | "END" 
<normal action> ::= "{" <sentences of Python> "}"
<list action> ::= <a list of Python>

B, BEGIN, E and END means the begin/end patterns, which are also seen in AWK. We can use BEGIN pattern for initialization and the other for post-processing. This is an example.

$ seq 100 | opy 'B:{a=0};{a+=F1};E:[a]'
5050

module

We can use any modules in opy. There are three kinds of ways for import of modules. In a list action, modules are imported automatically.

### obtaining the number of sin 1 ###
$ opy 'B:[math.sin(1)]'
0.8414709848078965
### calculation of sqrt(3*3 + 4*4) ###
$ opy 'B:[numpy.hypot(3,4)]'
5.0

Though it is crude implemntation, the module is imported when an NameError occurs. After that, opy evaluates the list again. Therefore, we must not write any procedure that has side effects in the list. Otherwise, a procedure runs twice.

We can import modules more safely with -m option.

$ opy -m numpy 'E:{print(numpy.pi)}'
3.141592653589793
### join module names with commas for multiple module import ###
$ opy -m math,numpy 'B:[math.e,numpy.e]'
2.718281828459045 2.718281828459045

 We can also use a begin pattern.

$ opy 'B:{import numpy};E:{print(numpy.pi)}'
3.141592653589793
$ opy 'B:{import numpy as np};E:{print(np.pi)}'
3.141592653589793

computational speed

I put less importance on computational speed than convenience.

repository

awesome example of use

It is available on ShellGei bot and some users enjoy using.

Thank you.



prev:Pythonをコマンドラインで使いたいのでopyというコマンドを作った next:自作コマンド(opy)をbrewとsnapに対応した