Fix the usage of Configuration class

[vdusek]

Our Configuration class currently contains many fields that are unused and/or undocumented. Review each field, decide whether it should be used, removed, or consolidated, and add missing doc strings.

[janbuchar]

Thanks for bringing this up! If you could write out the attributes in question here, we can discuss the appropriate steps together.

[vdusek]

The fields that are used:

• internal_timeout
• verbose_log (but it just sets the DEBUG level, and we have another field exactly for that)
• default_dataset_id
• default_key_value_store_id
• default_request_queue_id
• purge_on_start
• write_metadata
• persist_storage
• memory_mbytes
• storage_dir

The fields that are not used:

• default_browser_path
• disable_browser_sandbox
• persist_state_interval
• system_info_interval
• max_used_cpu_ratio
• available_memory_ratio
• chrome_executable_path
• headless
• xvfb

The inputKey field from JS is missing completely.

[janbuchar]

The fields that are not used:

• default_browser_path

In JS, this is used by PlaywrightLauncher

• disable_browser_sandbox

In JS, this is used by BrowserLauncher to add an argument to playwright/puppeteer

• chrome_executable_path

In JS, this is used by BrowserLauncher, and set in https://github.com/apify/apify-actor-docker/blob/master/node-playwright/Dockerfile or by the actor developer in environment variables

• xvfb

This may be set in https://github.com/apify/apify-actor-docker/blob/master/node-playwright/Dockerfile or by the actor developer in environment variables, but I don't think it actually has any effect 🤔 @vladfrangu do you know more?

• headless
  We should probably implement this - it's convenient for local debugging

• persist_state_interval
• system_info_interval

These should probably be consumed by LocalEventManager. Currently, the SDK passes that, which may not be optimal.

• max_used_cpu_ratio
• available_memory_ratio

These should probably be handled by LocalEventManager, SystemInfo or Snapshotter.

The inputKey field from JS is missing completely.

I must have missed that one. The only use case I can think of is when you want to debug something locally against multiple inputs and don't want to bother with renaming files... which is valid

[vdusek]

Agreed, thanks @janbuchar for the investigation. We should also note, that Configuration values should be overridable by component-specific values.

E.g.

config = Configuration(headless=False)
crawler = PlaywrightCrawler(config=config, headless=True)

then True is used for headless.

Also, we should split this issue into many smaller ones.