Tencent cloud client command line tool tccli main process analysis

environmental information

• python: 3.8.5
• tccli: 3.0.666.1

List only key code

Command line files

from tccli.main import main
if __name__ == '__main__':
    sys.argv[0] = re.sub(r'(-script\.pyw|\.exe)?$', '', sys.argv[0])
    sys.exit(main())

Entry function

main.py file

def main():
    cli_version = __version__.rsplit(".", 1)[0]
    if sdkVersion < cli_version:
        sys.stderr.write("Version is inconsistent, python sdk version:%s tccli version:%s" % (sdkVersion, __version__))
        return
    try:
        log.info("tccli %s" % ' '.join(sys.argv[1:]))
        CLICommand()()

main The function mainly calls CLICommand()().

command.CLICommand analysis

class CLICommand(BaseCommand):

    def __init__(self):
        self._command_map = None
        self._argument_map = None
        super(CLICommand, self).__init__()

    def __call__(self, args=None):
        if args is None:
            #  Save all parameters to  args 
            args = sys.argv[1:]

        if len(args) > 0 and args[0] == "as":
            args[0] = "autoscaling"

        command_map = self._get_command_map()
        parser = self._create_parser(command_map)

        #  Solve the interface version (--version)  and  tccli edition (--version) Field conflict 
        self._handle_service_version_argumnet(args, parser)
        self._handle_warning(args)

        parsed_args, remaining = parser.parse_known_args(args)
        return command_map[parsed_args.command](remaining, parsed_args)

command_map object

command_map It's a OrderedDict, When service by dnspod when , give the result as follows . To put it bluntly , Namely {"func_name": func} Such a dictionary , It is convenient to find the real function with the same name through the string .

The following example , Namely dnspod This service Under the Action.

Use tccli dnspod help You can see .

OrderedDict([('CreateDomain', <tccli.command.ActionCommand object at 0x7f6d7f0c16a0>), 
    ('CreateDomainAlias', <tccli.command.ActionCommand object at 0x7f6d7f0a5370>), 
    ('CreateDomainBatch', <tccli.command.ActionCommand object at 0x7f6d7f275970>), 
    ('CreateDomainGroup', <tccli.command.ActionCommand object at 0x7f6d7eae5310>), 
    ('CreateRecord', <tccli.command.ActionCommand object at 0x7f6d7f283160>), 
    ('CreateRecordBatch', <tccli.command.ActionCommand object at 0x7f6d7e7de430>), 
    ('DeleteDomain', <tccli.command.ActionCommand object at 0x7f6d7f2831c0>), ...])

parser object

This is actually CLIArgParser Class .

CLIArgParser come from argparser.py file

class CLIArgParser(BaseArgParser):

Its parent class BaseArgParser Inherited from argparse.ArgumentParser

class BaseArgParser(argparse.ArgumentParser):

parse_known_args() Method in BaseArgParser Rewrote .
Don't read it in detail , Let's take a direct look at the relevant results .

stay CLICommand Of __call__ Add print code under method .

print("parsed_args", parsed_args)
print("remaining", remaining)
sys.stdout.flush()

Output example 1

# tccli cls ModifyTopic help --detail
parsed_args Namespace(cli_input_json=None, cli_unfold_argument=False, command='cls', detail=True, endpoint=None, filter=None, generate_cli_skeleton=None, https_proxy=None, output=None, profile=None, region=None, role_arn=None, role_session_name=None, secretId=None, secretKey=None, service_version=None, timeout=None, token=None, use_cvm_role=False, waiter=None, warning=False)
remaining ['ModifyTopic', 'help']

Example 2

# tccli cls DescribeTopics --Filters '[ {"Key": "topicName", "Values": [ "test-log" ] }]'
parsed_args Namespace(cli_input_json=None, cli_unfold_argument=False, command='cls', detail=False, endpoint=None, filter=None, generate_cli_skeleton=None, https_proxy=None, output=None, profile=None, region=None, role_arn=None, role_session_name=None, secretId=None, secretKey=None, service_version=None, timeout=None, token=None, use_cvm_role=False, waiter=None, warning=False)
remaining ['DescribeTopics', '--Filters', '[ {"Key": "topicName", "Values": [ "test-log" ] }]']

Actually Action function , With doDescribeTopics For example

This function is defined in cls/cls_client.py The file of .

According to the following call , Its handle

• remaining As args Pass in
• parsed_args As parsed_globals Pass in

def doDescribeTopics(args, parsed_globals):
    ...
    model = models.DescribeTopicsRequest()
    model.from_json_string(json.dumps(args))
    start_time = time.time()
    while True:
        rsp = client.DescribeTopics(model)
        result = rsp.to_json_string()
    ...

It's easy to understand , Part of it is “ Global parameter ”, Part of it is “Action function ” Parameters of .

It will be “ Global parameter ” Turn into g_param, A print . give the result as follows ：

g_param {'filter': None, 'output': 'json', 'secretId': 'xxxxxxxxx', 'secretKey': 'xxxxxxxxxxx', 'token': None, 'role_arn': None, 'role_session_name': None, 'use_cvm_role': False, 'detail': False, 'profile': 'default', 'region': 'ap-beijing', 'endpoint': 'cls.tencentcloudapi.com', 'timeout': None, 'generate_cli_skeleton': None, 'cli_input_json': None, 'cli_unfold_argument': False, 'https_proxy': None, 'warning': False, 'waiter': None, 'command': 'cls', 'service_version': None, 'version': 'v20201016'}

model The object is v20201016/models.py in DescribeTopicsRequest Example . It mainly realizes _deserialize Method .

class DescribeTopicsRequest(AbstractModel):

from_json_string Method comes from abstract_model.py Medium AbstractModel class . yes DescribeTopicsRequest Parent class of .

And in the from_json_string The above _deserialize Method .

Here we are , Come to an end .

Invoke the sample

import sys
from tccli.main import main

args = ['tccli', 'cls', 'DescribeTopics',
    '--Filters',
    '[ {"Key": "topicName", "Values": [ "test-log" ] }]']

sys.argv = args
main()

Code above , In your own script , call tccli The function in .

Of course , Because of its main Always output results to stdout in , So if you want to save the result to a variable , You also need to write your own functions to handle it . This part can refer to python Save the standard output of the function to the variable problem .

format_output.py file

This file defines the operations related to the output format .

Response Class is the base class of other formats

__call__ Method self._format_response() Methods are implemented by subclasses , Such as JSONResult, TableResult,TextResult.

In the method , And settings stream Code for .

def __call__(self, command, response, stream=None):
    if stream is None:
        if not six.PY2:
            sys.stdout = io.TextIOWrapper(sys.stdout.buffer,
                                          encoding='utf-8')
        stream = sys.stdout
    try:
        self._format_response(command, response, stream)
    except IOError as e:
        pass
    finally:
        self._flush_stream(stream)

• First of all, judge whether there is stream Come in , In the current situation without modifying the source code , It can't be transmitted stream Parametric . So it must enter the logic of judgment
• stay python The version is not 2.x Under the circumstances , take sys.stdout Point to io.TextIOWrapper(sys.stdout.buffer,encoding='utf-8') Result .
  • This step , Mainly increased buffer effect ( At the same time, the encoding), Writing data will not be seen immediately , Examples are as follows
• And then stream Point to sys.stdout,

io.TextIOWrapper Effect demonstration

In [51]: sys.stdout =old_out

In [52]: print("hahha")
hahha

In [53]: print("hhehe")
hhehe

In [54]: sys.stdout = io.TextIOWrapper(sys.stdout.buffer,
    ...:                                               encoding='utf-8')
In [55]: print("hahha")
In [56]: print("haha")
In [57]: sys.stdout.flush()

hahha

haha

Use io.TextIOWrapper after ,print The result of will not show immediately , Instead, you need to call sys.stdout.flush() Function to see .

problem

In such as doDescribeRecordList In equal function , If used print() sentence , Then it needs to be implemented immediately sys.stdout.flush() Method , To see the output , Otherwise, only the corresponding function result will be output , I can't see what I added print() effect , Why is that ？

This reason is because later FormatOutput.output function , Changed the io flow , new io The flow is different from before . The verification code is as follows :

import sys
import io
import time

old_stdout = sys.stdout
sys.stdout = io.TextIOWrapper(old_stdout.buffer, encoding='utf-8')
print("I'm sys.stdout.")

#  Without this line , The script will report an error , I don't understand 
old_stdout = sys.stdout 
sys.stdout = io.TextIOWrapper(old_stdout.buffer, encoding='utf-8')
print(time.ctime(), "heheh")
print(time.ctime(),"hahah")
print("over")

function , The script above , You won't see I'm sys.stdout. Of , Because twice before and after io The flow is different .

But actually tccli in , Although it is also set twice sys.stdout, But it is not used as above old_stdout = sys.stdout This kind of sentence , Its script uses reload(sys).

• main.py Set once in the script sys.stdout
• format_output.py Set once in sys.stdout
  • When called by other functions

Try setting twice sys.stdout In different files , Not good either. .

Supported output formats

As in the code , Currently only supported 3 Ways of planting .

Think about it here , The possibility of customizing a format ？
Write a format by yourself , Put it in the source code , then , Specify this format yourself , Is it feasible ？

[[email protected] ~]# tccli --output table1 dnspod DescribeDomain --Domain hongyuan.com
usage: tccli [options] <command> <subcommand> [<subcommand> ...] [parameters]
To tccli help text, you can run:

  tccli help
  tccli configure help
  tccli service[cvm] help
  tccli service[cvm] action[RunInstances] help
tccli: error: argument --output: Invalid choice, valid choices are:

json                                     | text                                    
table                                   


Invalid choice: 'table1', maybe you meant:

  * table

Custom parsing format

There are parameter checks before starting , You need to cross this level first .

[[email protected] ~]# tccli --output json1 dnspod DescribeDomain --Domain hongyuan.com
usage: tccli [options] <command> <subcommand> [<subcommand> ...] [parameters]
To tccli help text, you can run:

  tccli help
  tccli configure help
  tccli service[cvm] help
  tccli service[cvm] action[RunInstances] help
tccli: error: argument --output: Invalid choice, valid choices are:

json                                     | text                                    
table                                   


Invalid choice: 'json1', maybe you meant:

  * json