Hi, guys. I'm really surprised about this > - (+) asserts look nicer and are more readable than if (something): > throw Exception()
I'm pretty sure that all the code I have encountered a way more readable using "if/else" or "try/except". But may be it is just me. Could you provide an example of code which is better with "assert"? - (+) asserts are especially good for cases like None exception - they > add more developer friendly messages when they will fail a few lines > below > with (for example) None has no property "dag". But it's ok if those get > optimised away. I think the best way to catch None is to ensure your code would fail conveniently. Like raising understandable Exception message, if you believe that should be a point of confusion. On Tue, 3 Dec 2019 at 16:22, Iuliia Volkova <[email protected]> wrote: > Hi everyone, I'm usually not write anything in this mail list, but this > theme something really strange > Exist offissial doc: > https://docs.python.org/3/reference/simple_stmts.html#the-assert-statement > > and there is a key information: Assert statements are a convenient way to > insert debugging assertions into a program. > > *Debugging. * - this is a key propose of asserts keyword. > > there is no any type of possible asserts that cannot be done with normal > Exceptions and Errors types that more explicit and detailed when 'assert' - > you have ValueError, TyperError and etc. what kind of problems must solved > DEBUG tools in production code that can be easily turned off on servers by > users? > > asserts used in tests and in process of debug code, not in production > > On Tue, Dec 3, 2019 at 6:47 PM Jarek Potiuk <[email protected]> > wrote: > > > We had a few discussions about using asserts in our code. I pasted some > > links below but wanted to extract a gist of it. > > > > Here are the comments summarised: > > > > - (+) asserts look nicer and are more readable than if (something): > > throw Exception() > > - (-) asserts can be optimized away with -O flag so we should not > based > > any real logic on having them > > - (+) asserts are good in cases that can happen in development but > > should "never happen" in reality > > - (+) asserts are especially good for cases like None exception - they > > add more developer friendly messages when they will fail a few lines > > below > > with (for example) None has no property "dag". But it's ok if those > get > > optimised away. > > > > We would like to discuss those points in community and have a community - > > driven decision on: > > > > 1) whether we should use asserts? > > 2) in which cases > > 3) in which cases we should NOT use asserts. > > > > J. > > > > The links here: > > > > Slack Discussion: > > https://apache-airflow.slack.com/archives/CCQ7EGB1P/p1575364664041300 > > > > Github threads: > > > > - https://github.com/apache/airflow/pull/6596#discussion_r352916409 > > - https://github.com/apache/airflow/pull/6596#discussion_r352914727 > > - > > https://github.com/apache/airflow/pull/3690#pullrequestreview-143376629 > > > > Stack overflow link for asserts: > > > > - https://stackoverflow.com/a/1838411/5691525 > > > > > > > > > > > > J. > > > > -- > > > > Jarek Potiuk > > Polidea <https://www.polidea.com/> | Principal Software Engineer > > > > M: +48 660 796 129 <+48660796129> > > [image: Polidea] <https://www.polidea.com/> > > > > > -- > _________ > > С уважением, Юлия Волкова > Тел. : +7 (911) 116-71-82 >
