generate a _deleted.txt file in CLI

[peterbe]

Part of #2224

When you use:

yarn tool delete Glossary/PNG

Generates (in files/en-us/_deleted.txt):

# This file is generated when you use the CLI tool to delete slugs.

/en-US/docs/Glossary/PNG                                     # Tue Jan 19 2021

And it works recursively and outside CONTENT_ROOT too:

yarn tool delete Glossary sv-SE -y -r

Generates (in files/sv-se/_deleted.txt):

# This file is generated when you use the CLI tool to delete slugs.

/sv-SE/docs/Glossary/404                                     # Tue Jan 19 2021
/sv-SE/docs/Glossary/Bandwidth                               # Tue Jan 19 2021
/sv-SE/docs/Glossary/IP_Adress                               # Tue Jan 19 2021
/sv-SE/docs/Glossary                                         # Tue Jan 19 2021

Now, this means we can start reading from this file, in the Deployer, before we proceed to upload stuff. This way, stuff will disappear much sooner from S3 and Elasticsearch.

[peterbe]

CC @escattone

[fiji-flo]

One nit. Bigger question is if we want to use a delete file approach or a "we gather everything and calculate the difference" approach.

[fiji-flo]

Suggested change

	const newLine = `${url.padEnd(60, " ")} # ${new Date().toDateString()}\n`;
	const newLine = `${url.padEnd(60, " ")}\t# ${new Date().toDateString()}\n`;

can we use a tab here? Easier to split and process.

[peterbe]

One nit. Bigger question is if we want to use a delete file approach or a "we gather everything and calculate the difference" approach.

Yeah, let's hold off on this PR.
The list of archived.txt URLs plus other strategies for the Elasticsearch publishing and stuff might just be enough. And if so, no need for this.

[escattone]

One nit. Bigger question is if we want to use a delete file approach or a "we gather everything and calculate the difference" approach.

Yeah, let's hold off on this PR.
The list of archived.txt URLs plus other strategies for the Elasticsearch publishing and stuff might just be enough. And if so, no need for this.

I had the same question as @fiji-flo, so I like that we're holding off for now and sleep on this some more.

[peterbe]

We decided that the Deployer can figure out what should be deleted based on the archived.txt instead.