Please read the naming conventions section of PEP 8, which explains the meaning of each of the styles. A brief overview of the most important parts:
Modules (and files) should use lowercase short names.
Class (and exception) names should use the
CapWordsconvention (also known as
Function and variable names should use lowercase letters, and words should be separated by underscores to improve readability (also called
To avoid conflicts with the standard library, you can add an underscore, such as
Leading lines such as
_dataapply to non-public methods and instance variables. Subclasses can use it. If you don’t use it in a subclass, use it like
__datain a superclass.
If there is a pair of
set_xmethods without additional parameters, please use the built-in
@propertydecorator to convert them to properties.
Constants should be similar to
When importing a function, try to avoid renaming it with
import asbecause it introduces cognitive overhead to keep track of another name. If the name conflicts, please use the package name as the namespace, such as
import schema, and use it as
If in doubt, follow existing conventions or check the style guide.
Test code excellence
Your code would be an example for others, and they might follow your approach. Therefore, both good and bad practices will be amplified.
In LISA, test code should be organized according to business logic, which means that the code should perform the purpose of the test like a test specification. The underlying logic should be implemented elsewhere, such as tools, functions, or private methods in test suites.
An example: Be careful when using
sleep! The only way to use sleep
is in polling mode. This means that you must wait for something with
regular inspections. In the inspection cycle, you can wait for a
reasonable period. Don’t wait for 10 seconds of sleep. This causes two
problems, 1) if it is too short, the case may fail; 2) if it is long
enough, it will slow down the running speed.
Please keep in mind that your code may be referred to by others.
How to write good code comments is a hot topic, and many best practices are also valuable. Here are some highlights.
Do not repeat the code logic. Code comments are always in the same place as the code, which is different from metadata. Do not repeat
if/elsestatement like “if … else …”, do not repeat the content that already exists in the log string and exception message, do not repeat what can be clearly seen from the variable name.
Record business logic. Code logic is more detailed than business logic. Some complex code logic may not be intuitive for understanding business logic. Code comments can help summarize complex code logic.
Record trick things. We cannot avoid writing tricky code. For example, magic numbers, special handling of the Linux version, or other content.
Provide regular expression examples. LISA uses many regular expressions to parse command output. It is simple and useful, but it may not match. When you need to create or update a regular expression, it needs to check the sample for regression. These examples also help to understand what the expression does.
The commit message is used to explain why this change was made. The code comments describe the current state. The commit message describes the reason for the change. If you think the content is also suitable for writing in the code, please write it as a code comment.
The log has two purposes, 1) display progress, and 2) troubleshoot.
To show progress, the log should be simple and logical. To troubleshoot, it requires more detailed information. These two goals sound contradictory, but they can be achieved through different INFO and DEBUG levels. LISA always enables the DEBUG level in the log file, while the INFO level is the default setting on the console.
In LISA, when writing log lines in the code, it’s recommended to consider what the test runner needs to know, instead of what the developer needs to know, which should be done in code comments.
DEBUG level log should provide the correct level detail. The only way to write at the “correct level” is to use it from the beginning.
When writing code, please keep using and improving the log. If you need to debug step by step, it means you need to improve the log. If you don’t understand the meaning of the log, others may not as well, so please optimize the log at DEBUG level. In addition, if you find duplicate information, please merge it.
INFO level log should be like a story, to illustrate what happened.
Even if the whole process goes smoothly, this is what you want to know every time. It should be friendly so that new users can understand what is going on. It should be as little as possible. It should tell the user to wait before performing a long operation.
WARNING level logs should be avoided.
The warning message indicates that it is important, but there is no need to stop. But in most cases, you will find that it is either not as important as the information level, or it is so important to stop running.
At the time of writing, there are 3 warning messages in LISA. After review, I converted them all into information or error level. There is only one left, and it is up to the user to suppress errors.
ERROR level log should be reviewed carefully.
Error level logs can help identify potential problems. If there are too many error level logs, it will hide the actual problem. When it goes smoothly, there should be no error level logs. According to experience, 95% of successful runs should not contain any error level logs.
By reading the log, you should be able to understand the progress without having to look at the code. And logs describe business logic, not code logic. A bad example, “4 items found: [a , b , c]”, should be “found 4 channels, unique names: [a, b, c]”.
Make each log line unique in the code. If you must check where the log is printed in the code. We can quickly find the code by searching. A bad example,
log.info("received stop signal"), should be
log.info("received stop signal in lisa_runner").
Do not repeat similar lines in succession. It is worth adding logic and variables to reduce redundant logs.
Reduce log lines. If two lines of logs always appear together, merge them into one line. The impact of log lines on readability is much greater than the length of the log.
Associate related logs through shared context. In the case of concurrency, this is very important. A bad example, “cmd: echo hello world”, “cmd: hello world” can be “cmd: echo hello world”, “cmd: hello world”.
There are two kinds of error messages in LISA. The first is an error message, and it does not fail. It will be printed as stderr and will be more obvious when the test case fails. The second is a one-line message in the failed test case. This section applies to two of them, but the second one is more important because we want it to be the only information that helps understand the failed test case.
In LISA, failed, skipped, and some passed test cases have a message. It specifies the reason the test case failed or skipped. Through this message, the user can understand what will happen and can act. Therefore, this message should be as helpful as possible.
The error message should include what happened and how to resolve it. It may not be easy to provide all the information for the first time, but guesswork is also helpful. At the same time, the original error message is also useful, please don’t hide it.
“The subscription ID [aaa] could not be found, please make sure it exists and is accessible by the current account”. A bad example, “The subscription ID [aaa] could not be found”. This bad example illustrates what happened, but there is no suggestion.
“The vm size [aaa] could not be found on the location [bbb]. This may be because the virtual machine size is not available in this location”. A bad example, “The vm size [aaa] could not be found on the location [bbb]”. It explains what happened, but it does not provide a guess at the root cause.
Assertions are heavily used in test code. Assertions are a simple pattern of “if some checks fail, raise an exception”.
The assertion library includes commonly used patterns and detailed error
messages. LISA uses
assertpy as a standard assertion library, which
provides Pythonic and test-friendly assertions.
When writing the assertion,
Put the actual value in
assert_thatto keep the style consistent, and you can compare it with multiple expected values continuously.
Assertions should be as comprehensive as possible, but do not repeat existing checks. For example,
assert_that(str1).is_equal_to('hello')is enough, no need like
Add a description to explain the business logic. If a malfunction occurs, these instructions will be displayed. For example,
assert_that(str1).described_as('echo back result is unexpected').is_equal_to('hello')is better than
Try to use native assertions instead of manipulating the data yourself.
assert_that(vmbuses).is_length(6)is better than
assert_that(len(vmbuses)).is_equal_to(6). It is simpler and the error message is clearer.
Don’t forget to use powerful collection assertions. They can compare ordered list by
contains(actual value is superset),
is_subset_of(actual value is subset), and others.
Test failure is a common phenomenon. Therefore, perform troubleshooting frequently. There are some useful ways to troubleshoot failures. In the list below, the higher items are better than the lower items because of its lower cost of analysis.
Single line message. A one-line message is sent with the test result status. If this message clearly describes the root cause, no other digging is necessary. You can even perform some automated actions to match messages and act.
Test case log. LISA provides a complete log for each run, which includes the output of all test cases, all threads, and all nodes. This file can be regarded as the default log, which is easy to search.
Other log files. Some original logs may be divided into test cases. After finding out the cause, it is easier to find out. But it needs to download and browse the test result files.
Reproduce in the environment. It is costly but contains most of the original information. But sometimes, the problem cannot be reproduced.
In LISA, test cases fail due to exceptions, and exception messages are treated as single-line messages. When writing test cases, it’s time to adjust the exception message. Therefore, after completing the test case, many errors will be explained well.
The documentation is the opportunity to make things clear and easy to maintain. A longer document is not always a better document. Each kind of documentation has its own purpose. Good technical documentation should be useful and accurate.
Tips for non-native English speakers by non-native English speakers
Today, there are a lot of great tools to help you create high-quality English documents. If writing in English is challenging, please try the following steps:
Read our documentations.
Write in your language first.
Convert the English version back to your language and check. If it doesn’t make sense after translating back, it means the sentence is too complicated. Make it simpler, and then start from step 1 again.
Once satisfied, you can use Microsoft Editor to further refine the grammar and wordings.