instance Default WidgetFileSettings vanished

[thielema]

After a cabal update I get the following type error when compiling hledger-web:

    • No instance for ‘Data.Default.Internal.Default
                         WidgetFileSettings’
        arising from a use of ‘def’

Did something change in one of the yesod packages?

[jezen]

The file where this instance is defined hasn't changed in the best part of eight years. I don't see any changes to Yesod that would have caused this in the past half year either. Could the change be somewhere else, like in hledger?

[thielema]

On Thu, 24 Oct 2024, Jezen Thomas wrote: The file where this instance is defined hasn't changed in the best part of eight years. I don't see any changes to Yesod that would have caused this in the past half year either. Could the change be somewhere else, like in hledger?

I assumed it could be caused by Cabal choosing two different versions of the same yesod package at different places. But Cabal has not shown a warning about such a conflict. Hm.

[jezen]

Could it perhaps be related to this?

#1848 (comment)

[thielema]

On Thu, 24 Oct 2024, Jezen Thomas wrote: Could it perhaps be related to this? #1848 (comment)

That one looks different. I assumed my problem could have to do with an orphan instance, but "instance Default WidgetFileSettings" is defined right after "WidgetFileSettings" in Yesod.Default.Util.

[thielema]

On Thu, 24 Oct 2024, thielema wrote: I assumed my problem could have to do with an orphan instance, but "instance Default WidgetFileSettings" is defined right after "WidgetFileSettings" in Yesod.Default.Util.

I think I found the problem. yesod imports package data-default-class and defines instance Data.Default.Class.Default WidgetFileSettings, whereas hledger expects instance Data.Default.Default WidgetFileSettings from package data-default. Formerly, data-default just re-exported Default class from data-default-class, but in data-default-0.8.0.0 there seems to be a custom definition. data-default-0.8.0.0 was released yesterday. I do not know whether this means that data-default-class becomes deprecated or what the relation between data-default and data-default-class is intended to be.

[thielema]

The author of data-default-class tells, that in future data-default should be preferred:
mauke/data-default#29

[jezen]

Thank you for investigating. It seems (and correct me if I'm wrong) that there's nothing really to do here though, so I will close this issue.

[thielema]

On Sat, 26 Oct 2024, Jezen Thomas wrote: Thank you for investigating. It seems (and correct me if I'm wrong) that there's nothing really to do here though, so I will close this issue.

I think you can replace data-default-class by data-default to become future-proof.