strconv: raise error if not a number

[TheBrainScrambler]

Hello, I've just discovered the language and have read the docs. Now I want to practice, so I'm writing a little guessing game where you enter a number between 1 and 100.
I need to convert the user's input from a string to an int. Reading the docs, I've found https://modules.vlang.io/strconv.html#atoi which does what I want. BUT ... not exactly.
If I enter some random characters in my program, strconv.atoi will just convert it to 0 instead of rising an error where it says that the string can't be converted.
I think this is incorrect behavior, the strconv methods should be Optionals, or at least provide methods which implements this mechanisms and are Optionals.

TL;DR
Expected behavior: If I do strconv.atoi("blablah") it should let us handle the error with an Optional
Current behavior: It just returns 0 ...

[Delta456]

What's your say on this? @penguindark

[medvednikov]

number := os.input('enter number').int()

[medvednikov]

But I agree, atoi() should return ?int

[TheBrainScrambler]

"number := os.input('enter number').int()"
This also returns 0 if I enter characters instead of returning an error

[Delta456]

if we add an optional for so str.int() will return 0 or error if it's a not a valid number right? @medvednikov

[TheBrainScrambler]

@Delta456 No, it just returns 0 if the number is invalid, like strconv.atoi. Test yourself by running number := os.input('enter number').int() and printing number. Enter "Blablah" for example, it will print you 0.

[Delta456]

@TheBrainScrambler I meant if we will add an optional so what will it return.

[TheBrainScrambler]

@Delta456 I don't understand, the optional should return 0 if it isn't an error ? What ?

I also discovered by using the cli module that there is the same behavior: it will convert my entry to 0 if it isn't a valid number, instead of returning an error. It is probably using strconv.atoi or the .int() method

[OpenSourceAnarchist]

Yeah I'm for making pretty much all user-interaction functions return an optional type. There should be no overlap between valid returns (like 0 in this case) and invalid returns (also 0 in this case). This seems like the perfect use-case of optionals.

[penguindark]

@medvednikov @Delta456
atoi in C act like: This function returns the converted integral number as an int value. If no valid conversion could be performed, it returns zero
In V it has the same behaviour.
Adding an optional can be done but I'm not sure if it is needed.

[OpenSourceAnarchist]

@penguindark C does not have optionals, so returning 0 is the best solution within the constraints of the language. V has a feature that would lend itself perfectly for parsing "iffy" data (unknown strings) with the certainty that whatever is returned by the function, if not an error, is precisely what the developer wants. Without optionals, there is no way to know if the user typed 0 or if they typed nonsense.

[JalonSolov]

It would perhaps be better to create a different method with a new name, instead of changing the way atoi works.

The official documentation for the parameter to atoi (in C, C++, etc.) is:

C-string beginning with the representation of an integral number.

The key part (that I bolded) is "beginning with".

This also means that atoi will return 5 if you pass it a string of "5ive", instead of giving an error.

So having a separate strict version of atoi would be ok, but changing the way atoi works would just make porting existing code harder.

[penguindark]

I agree with JalonSolov
Add another method can be the best solution.

[TheBrainScrambler]

I don't know if the strconv.atoi and/or .int() come from C or V. Maybe just keep atoi for C compatibility by mentionning that in the docs and just modify the int method ... It would also be interesting to know what the cli module uses to convert strings to int

[JalonSolov]

Looks like .int()... example line:

return flag.value.int()

[TheBrainScrambler]

Then I think the .int() method should be modified to return an Optional and keep strconv.atoi for C compatibility.

This should also be marked as a bug. I can't properly use the CLI library because if an user enter some string on an int flag, it won't raise an error and instead return 0. But 0 is also a valid entry so I can't check if the entry is 0 ...

[medvednikov]

Yeah good point

[mcastorina]

strconv.atoi returns ?int (V 0.2.2 6e589e2)

Can this issue be closed?

[medvednikov]

Indeed it can now :)

[thomas-mangin]

As a newcomer to the language, I was extremely surprised to find that .int() or .u32() did not use optional and require an or block. I had to find this issue to figure out what to do to make sure that invalid input would not result in a zero value.

the introduction of strconv.atoi did allow for both use cases (thank you) but is IMHO un-intuitive. strconv.atoi should have followed the behaviour of C (return 0) when the built-in should have provided an optional and be "safe by default".

As this issue is not closed, perhaps the behaviour can still be changed and vfmt used to invert the use of one with the other. If so, atou32, and friends may need to be added to vlib to allow a full substitution.

[prantlf]

Indeed, atoi in C returns 0 in case of failure, which makes it useless for obtain numbers including 0. Instead of atoi, strtol has been recommended for a long time. It still returns 0 in case of a failure, but it also sets errno to EINVAL.

So, .int() in V ended up like atoi in C - next to useless. The equivalent of strtol in C was named atoi in V. How ironic :-)

[Wajinn]

Seems like those who were dealing with C compatibility, would have already encountered or adapted to ! from V's strconv.atoi and strconv.parse_int. And if they ran over to using .int(), then maybe split it into 2 new functions, where one gives zero and the other is optional.

[prantlf]

You could add a rule to v vet, which would warn about the usage of .int() and recommend replacing it with strconv.atoi(). It's avoid mistakes from the omitting the error handling.

[Wajinn]

At least updating the documentation (both the V docs and STL docs) to provide more clarity, seems like the best move, regardless of what else is decided later. In the V docs, as well as links to strconv and os, possibly also mention strconv.atoi, strconv.parse_int, and os.input (or others) too. This way, newbies know to look there. In the STL docs, maybe add example usage for those mentioned.

[thomas-mangin]

@prantlf imho the built-in should be working so that users do not need any import and the library should have kept the C api. Just my view.

[medvednikov]

@Wajinn is correct. That was my goal: have a simple to use method that returns 0 on error, and a more advanced function that returns a Result.