Configuration Definition Refactor, Phase 1

[isaacs]

This is the first step in refactoring our configuration definition, flattening, and documentation.

Still TODO:

• Remove all use of npm.flatOptions in lib/cli, other than passing to dependencies. (@wraithgar is working on this in another PR)
• Audit every dependency's use of every option key, so that we know which things are used in which ways. (And so that we can stop flattening options just for our own internal usage, since we can access config.get() which is fine.)
• Categorize config keys into groups for the purpose of easily documenting which configs are used by which commands.
• Identify (ie, in the config definition) which flatOptions are affected by which config fields.
• Move back to a single place where flatOptions are flattened (so that logic isn't spread across multiple config definitions in cases like saveType)

Furthermore, explore whether the flatOptions dirty flag should be "you need to regenerate all of these", or a dirty flag for the individual fields that might have changed, or just an object full of getters for each one. The current approach (generate flat options if missing, throw away when config.set() changes anything) is naive, but simple and probably performant enough.

[isaacs]

This is ready to review for basic structure and approach, but there are a few blockers to landing it:

• Should remove all use of npm.flatOptions from lib/*.js, except for passing to other deps.
• Finish audit of config field usage (this is also turning up a few errors which will be fixed shortly).
• Land changes to node_modules/@npmcli/config upstream and publish a major, so that when we pull it in, it's a "clean fork" from the published module before making any other changes.

[isaacs]

This is the first step in refactoring our configuration definition, flattening, and documentation.

Still TODO:

• Remove all use of npm.flatOptions in lib/cli, other than passing to dependencies. (@wraithgar is working on this in another PR)
• Audit every dependency's use of every option key, so that we know which things are used in which ways. (And so that we can stop flattening options just for our own internal usage, since we can access config.get() which is fine.)
• Categorize config keys into groups for the purpose of easily documenting which configs are used by which commands.
• Identify (ie, in the config definition) which flatOptions are affected by which config fields.
• Move back to a single place where flatOptions are flattened (so that logic isn't spread across multiple config definitions in cases like saveType)

Furthermore, explore whether the flatOptions dirty flag should be "you need to regenerate all of these", or a dirty flag for the individual fields that might have changed, or just an object full of getters for each one. The current approach (generate flat options if missing, throw away when config.set() changes anything) is naive, but simple and probably performant enough.

[isaacs]

This is ready to review for basic structure and approach, but there are a few blockers to landing it:

• Should remove all use of npm.flatOptions from lib/*.js, except for passing to other deps.
• Finish audit of config field usage (this is also turning up a few errors which will be fixed shortly).
• Land changes to node_modules/@npmcli/config upstream and publish a major, so that when we pull it in, it's a "clean fork" from the published module before making any other changes.

[isaacs]

This is the first step in refactoring our configuration definition, flattening, and documentation.

Still TODO:

• Remove all use of npm.flatOptions in lib/cli, other than passing to dependencies. (@wraithgar is working on this in another PR)
• Audit every dependency's use of every option key, so that we know which things are used in which ways. (And so that we can stop flattening options just for our own internal usage, since we can access config.get() which is fine.)
• Categorize config keys into groups for the purpose of easily documenting which configs are used by which commands.
• Identify (ie, in the config definition) which flatOptions are affected by which config fields.
• Move back to a single place where flatOptions are flattened (so that logic isn't spread across multiple config definitions in cases like saveType)

Furthermore, explore whether the flatOptions dirty flag should be "you need to regenerate all of these", or a dirty flag for the individual fields that might have changed, or just an object full of getters for each one. The current approach (generate flat options if missing, throw away when config.set() changes anything) is naive, but simple and probably performant enough.

[isaacs]

This is ready to review for basic structure and approach, but there are a few blockers to landing it:

• Should remove all use of npm.flatOptions from lib/*.js, except for passing to other deps.
• Finish audit of config field usage (this is also turning up a few errors which will be fixed shortly).
• Land changes to node_modules/@npmcli/config upstream and publish a major, so that when we pull it in, it's a "clean fork" from the published module before making any other changes.

[isaacs]

This is the first step in refactoring our configuration definition, flattening, and documentation.

Still TODO:

• Remove all use of npm.flatOptions in lib/cli, other than passing to dependencies. (@wraithgar is working on this in another PR)
• Audit every dependency's use of every option key, so that we know which things are used in which ways. (And so that we can stop flattening options just for our own internal usage, since we can access config.get() which is fine.)
• Categorize config keys into groups for the purpose of easily documenting which configs are used by which commands.
• Identify (ie, in the config definition) which flatOptions are affected by which config fields.
• Move back to a single place where flatOptions are flattened (so that logic isn't spread across multiple config definitions in cases like saveType)

Furthermore, explore whether the flatOptions dirty flag should be "you need to regenerate all of these", or a dirty flag for the individual fields that might have changed, or just an object full of getters for each one. The current approach (generate flat options if missing, throw away when config.set() changes anything) is naive, but simple and probably performant enough.

[isaacs]

This is ready to review for basic structure and approach, but there are a few blockers to landing it:

• Should remove all use of npm.flatOptions from lib/*.js, except for passing to other deps.
• Finish audit of config field usage (this is also turning up a few errors which will be fixed shortly).
• Land changes to node_modules/@npmcli/config upstream and publish a major, so that when we pull it in, it's a "clean fork" from the published module before making any other changes.

[isaacs]

This is the first step in refactoring our configuration definition, flattening, and documentation.

Still TODO:

• Remove all use of npm.flatOptions in lib/cli, other than passing to dependencies. (@wraithgar is working on this in another PR)
• Audit every dependency's use of every option key, so that we know which things are used in which ways. (And so that we can stop flattening options just for our own internal usage, since we can access config.get() which is fine.)
• Categorize config keys into groups for the purpose of easily documenting which configs are used by which commands.
• Identify (ie, in the config definition) which flatOptions are affected by which config fields.
• Move back to a single place where flatOptions are flattened (so that logic isn't spread across multiple config definitions in cases like saveType)

Furthermore, explore whether the flatOptions dirty flag should be "you need to regenerate all of these", or a dirty flag for the individual fields that might have changed, or just an object full of getters for each one. The current approach (generate flat options if missing, throw away when config.set() changes anything) is naive, but simple and probably performant enough.

[isaacs]

This is ready to review for basic structure and approach, but there are a few blockers to landing it:

• Should remove all use of npm.flatOptions from lib/*.js, except for passing to other deps.
• Finish audit of config field usage (this is also turning up a few errors which will be fixed shortly).
• Land changes to node_modules/@npmcli/config upstream and publish a major, so that when we pull it in, it's a "clean fork" from the published module before making any other changes.

[isaacs]

This is the first step in refactoring our configuration definition, flattening, and documentation.

Still TODO:

• Remove all use of npm.flatOptions in lib/cli, other than passing to dependencies. (@wraithgar is working on this in another PR)
• Audit every dependency's use of every option key, so that we know which things are used in which ways. (And so that we can stop flattening options just for our own internal usage, since we can access config.get() which is fine.)
• Categorize config keys into groups for the purpose of easily documenting which configs are used by which commands.
• Identify (ie, in the config definition) which flatOptions are affected by which config fields.
• Move back to a single place where flatOptions are flattened (so that logic isn't spread across multiple config definitions in cases like saveType)

Furthermore, explore whether the flatOptions dirty flag should be "you need to regenerate all of these", or a dirty flag for the individual fields that might have changed, or just an object full of getters for each one. The current approach (generate flat options if missing, throw away when config.set() changes anything) is naive, but simple and probably performant enough.

[isaacs]

This is ready to review for basic structure and approach, but there are a few blockers to landing it:

• Should remove all use of npm.flatOptions from lib/*.js, except for passing to other deps.
• Finish audit of config field usage (this is also turning up a few errors which will be fixed shortly).
• Land changes to node_modules/@npmcli/config upstream and publish a major, so that when we pull it in, it's a "clean fork" from the published module before making any other changes.

[isaacs]

This is the first step in refactoring our configuration definition, flattening, and documentation.

Still TODO:

• Remove all use of npm.flatOptions in lib/cli, other than passing to dependencies. (@wraithgar is working on this in another PR)
• Audit every dependency's use of every option key, so that we know which things are used in which ways. (And so that we can stop flattening options just for our own internal usage, since we can access config.get() which is fine.)
• Categorize config keys into groups for the purpose of easily documenting which configs are used by which commands.
• Identify (ie, in the config definition) which flatOptions are affected by which config fields.
• Move back to a single place where flatOptions are flattened (so that logic isn't spread across multiple config definitions in cases like saveType)

Furthermore, explore whether the flatOptions dirty flag should be "you need to regenerate all of these", or a dirty flag for the individual fields that might have changed, or just an object full of getters for each one. The current approach (generate flat options if missing, throw away when config.set() changes anything) is naive, but simple and probably performant enough.

[isaacs]

This is ready to review for basic structure and approach, but there are a few blockers to landing it:

• Should remove all use of npm.flatOptions from lib/*.js, except for passing to other deps.
• Finish audit of config field usage (this is also turning up a few errors which will be fixed shortly).
• Land changes to node_modules/@npmcli/config upstream and publish a major, so that when we pull it in, it's a "clean fork" from the published module before making any other changes.

[isaacs]

This is the first step in refactoring our configuration definition, flattening, and documentation.

Still TODO:

• Remove all use of npm.flatOptions in lib/cli, other than passing to dependencies. (@wraithgar is working on this in another PR)
• Audit every dependency's use of every option key, so that we know which things are used in which ways. (And so that we can stop flattening options just for our own internal usage, since we can access config.get() which is fine.)
• Categorize config keys into groups for the purpose of easily documenting which configs are used by which commands.
• Identify (ie, in the config definition) which flatOptions are affected by which config fields.
• Move back to a single place where flatOptions are flattened (so that logic isn't spread across multiple config definitions in cases like saveType)

Furthermore, explore whether the flatOptions dirty flag should be "you need to regenerate all of these", or a dirty flag for the individual fields that might have changed, or just an object full of getters for each one. The current approach (generate flat options if missing, throw away when config.set() changes anything) is naive, but simple and probably performant enough.

[isaacs]

This is ready to review for basic structure and approach, but there are a few blockers to landing it:

• Should remove all use of npm.flatOptions from lib/*.js, except for passing to other deps.
• Finish audit of config field usage (this is also turning up a few errors which will be fixed shortly).
• Land changes to node_modules/@npmcli/config upstream and publish a major, so that when we pull it in, it's a "clean fork" from the published module before making any other changes.