--write-metadata shouldn't override metadata options in configuration file

[Vrihub]

I have set metadata options in my configuration file so that gallery-dl always downloads metadata in a custom directory, and it works OK.

However, if by mistake I also add --write-metadata on the command line, it seems that options in the configuration files are ignored; especially: metadata are not written anymore in my custom directory but in the default directory (=the same where images are saved).

[Hrxn]

Uh, no, that's basically how all CLI programs behave. If you have set the same option (i.e. write-metadata) in both your config file and on the command-line, the command-line always overwrites the config file.

[Vrihub]

Uh, no, that's basically how all CLI programs behave...

Of course, but this holds if the option on CL and the one in the config file are the same.

The problem here is that --write-metadata, which is (should be ?) just a switch to decide whether to write metadata or not, overrides another option, i.e. metadata.directory in config. It shouldn't have any effect here (since metadata writing is already configured in the file) and it shouldn't reset other metadata options to the default value.

[mikf]

which is (should be ?) just a switch to decide whether to write metadata or not

Well, it's not, or at least that's not how it's implemented. --write-metadata sets a global postprocessors setting, overriding all other post-processor settings in the process.

What you could do is define your metadata settings in a secondary config file and only load it with -c/--config if you want to enable writing metadata:

metadata.conf
-------------

{ "postprocessors": [
    {
        "name": "metadata",
        "other": "settings"
    }
] }

$ gallery-dl -c metadata.conf URL

[Vrihub]

Well, it's not, or at least that's not how it's implemented. --write-metadata sets a global postprocessors setting, overriding all other post-processor settings in the process.

Uhm, are there other global "switch" options that behave in this way?
At first glance, I would check --write-tags and --zip.

If --write-metadata is the only one, I would insist that it should be fixed (i.e. have no effect if the config file already contains any postprocessors.metadata options).

If you want to keep this behaviour instead, I think it should be documented in docs/configuration.rst and in the man page (I can try to come up with a patch).

[mikf]

Uhm, are there other global "switch" options that behave in this way?

Yes, almost all of them.
Most command-line switches are implemented by setting a global config setting at the base level that "overwrites" all config options with the same name.

It usually makes sense for "simple" values like HTTP timeouts, proxy servers, etc. that a command-line switch takes precedence over a config setting, but for complex/composite stuff like a list of post processor objects it does seem weird, I agree.

I'll see if I can improve the current documentation, but I'd also gladly accept a patch from you.

[Vrihub]

I guess this issue can be closed