How to migrate away from 'nameof'?

[zapta]

We have a code similar to the one below where we use nameof to convert arg names to strings. The advantage is that the ide tools are aware of the relation between the function args and nameof's args (e.g. when renaming or linting).

We tried to upgrade to the latest varname version and get a warning that we should use argname() instead but it seems to require the name of the arg a a string which defeats the reason we use varname.

Are we missing anything? What is the use case of argname() if the user requires to provide the arg name string?

def cli(
    cmd_ctx: click.core.Context,
    # Options:
    ftdi_install: bool,
    ftdi_uninstall: bool,
    serial_install: bool,
    serial_uninstall: bool,
):
    """Implements the drivers command."""

    # User should select exactly on of these operations.
    cmd_util.check_exactly_one_param(
        cmd_ctx,
        nameof(ftdi_install, ftdi_uninstall, serial_install, serial_uninstall),
    )

[pwwang]

What do you think this will return:

nameof(ftdi_install, ftdi_uninstall, serial_install, serial_uninstall)?

[zapta]

I think ["ftdi_install", "ftdi_uninstall", 'serial_install", "serial_uninstall"] That is, nameof maps arg references to arg names.

…

On Sat, Dec 21, 2024 at 8:04 PM pwwang ***@***.***> wrote: What do you think this will return: nameof(ftdi_install, ftdi_uninstall, serial_install, serial_uninstall)? — Reply to this email directly, view it on GitHub <#117 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAQVMQK2ELRSO4GDTNUVGZL2GY23TAVCNFSM6AAAAABUBHXWU2VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDKNJYGMZDEMBWGI> . You are receiving this because you authored the thread.Message ID: ***@***.***>

[pwwang]

So then why not just use ["ftdi_install", "ftdi_uninstall", 'serial_install", "serial_uninstall"] directly?

[zapta]

Because the IDE and tools like linters will not be aware of the association
between the arg ftdi_install and the string 'ftdi_install'. For
example if we rename the arg of if the string has a typo.

[pwwang]

I don't get it. If you change, say, ftdi_install to ftdi_install2, then what will the IDE/linters complain about if you do:

def cli(
    cmd_ctx: click.core.Context,
    # Options:
    ftdi_install2: bool,
    ftdi_uninstall: bool,
    serial_install: bool,
    serial_uninstall: bool,
):
    """Implements the drivers command."""

    # User should select exactly on of these operations.
    cmd_util.check_exactly_one_param(
        cmd_ctx,
        ["ftdi_install2", "ftdi_uninstall", "serial_install", "serial_uninstall"),
    )

[zapta]

The tools will not complain that there is no arg called ftdi_install3, but with nameof() they will. Same if I use an IDE refactoring tool to rename the arg ftdi_uninstall to ftdi_install4, it will not rename also the string 'ftdi_uninstall', but with nameof() it will.

def cli(
    cmd_ctx: click.core.Context,
    # Options:
    ftdi_install2: bool,
    ftdi_uninstall: bool,
    serial_install: bool,
    serial_uninstall: bool,
):
    """Implements the drivers command."""

    # User should select exactly on of these operations.
    cmd_util.check_exactly_one_param(
        cmd_ctx,
        ["ftdi_install3", "ftdi_uninstall", "serial_install", "serial_uninstall"),
    )

[pwwang]

I am still confused. Why did you put "ftdi_install3" instead of "ftdi_install2"?

[zapta]

To demonstrate a mismatch (e.g. a typo) between the arg name and the string. If instead of "ftdi_install3" I use nameof(ftdi_install3), then the tools detect the error statically, in the editor.

[pwwang]

I see what you meant. But I am not convinced that nameof should be used for this purpose. But if you do want it to be used this way, you can wrap a version of nameof by yourself using argname:

from varname import argname

def my_nameof(*args):
  return argname('*args')

a = b = 1
my_nameof(a, b)  # ('a', 'b')

[zapta]

You can look at the code in the link below. All the places that argument fix is mentioned are associated with each other and if I rename one they all change. With strings they are isolated from each other and the tools don't know about the intended association.

https://github.com/FPGAwars/apio/blob/ca5dc994ff7f66b1f752db57bfad3b97a0eb0772/apio/commands/packages.py#L222

[pwwang]

What you referred to as "associated" or "isolated" is merely due to the help of the IDE, rather than being based on the code itself.

[zapta]

Thanks! I will give it a try. I believe that they are associated by the semantics of python, because they are two references to the same semantic construct. On the other hand, two variables x in different functions are considered independent of each other. The tools just follow the semantic of the language.

…

On Sat, Dec 21, 2024 at 10:50 PM pwwang ***@***.***> wrote: What you referred to as "associated" or "isolated" is merely due to the help of the IDE, rather than being based on the code itself. — Reply to this email directly, view it on GitHub <#117 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAQVMQLLOWDIE5YMM4MCB3D2GZOM3AVCNFSM6AAAAABUBHXWU2VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDKNJYGM2TENRRGY> . You are receiving this because you authored the thread.Message ID: ***@***.***>

[pwwang]

You are correct, and that is the foundation of the IDE. I won’t argue about code style. However, nameof has been deprecated due to its redundancy with argname, and it's quite straightforward to implement it using argname (as I demonstrated in the above example). Another reason for the deprecation is that many people tend to use it incorrectly. In many cases, literal strings are sufficient.

[zapta]

I implemented my_nameof and it works, thanks.

I just had to rename args to _args to make the tools happy since from their perspective it's an unused argument.