Isaacs/update libnpmetc stuff by isaacs · Pull Request #968 · npm/cli

isaacs · 2020-03-03T08:12:36Z

Refactor of mikemimik/update-libnpmhook branch from pairing earlier today. Removes a lot of figgy-pudding and blue birds.

Commands updated with a new lib:

hook
access
team
org
search

Other random stuff:

cache.js
deprecate.js
fund.js
logout.js
owner.js
ping.js
star.js
stars.js
version.js
view.js
whoami.js

Utils:

utils/otplease.js
utils/git.js
utils/pulse-till-done.js
utils/read-user-info.js

Still todo:

publish (libnpmpublish)
unpublish (also libnpmpublish)
pack (should leverage pacote more than it does)
audit (should use @npmcli/arborist)
shrinkwrap (should use @npmcli/arborist)
update (should use @npmcli/arborist)
outdated (should use @npmcli/arborist)
ci (should use @npmcli/arborist)

isaacs · 2020-03-03T17:14:34Z

Note: this closes #935 as well.

mikemimik

Some discussion points, @isaacs let me know if you want me to handle any of this minutia.

mikemimik · 2020-03-04T16:54:30Z

 const readUserInfo = require('./read-user-info.js')

+const isOtpError = err =>
+ err.code === 'EOTP' || err.code === 'E401' && /one-time pass/.test(err.body)


Silly question, does this need brackets to evaluate correctly?
Suggested change
err.code==='EOTP'||err.code==='E401'&&/one-timepass/.test(err.body)
err.code==='EOTP'||(err.code==='E401'&&/one-timepass/.test(err.body))

It does not. && has higher priority than ||. In most expressions of boolean calculus, you can think of and (or any intersection operator) like multiplication or division, and or (or any union operator) like addition or subtraction. So a || b && c || d is like a + b * c + d, so it's like a + (b * c) + d and thus a || (b && c) || d.

mikemimik · 2020-03-05T03:46:56Z

 strictSSL: npm.config.get('strict-ssl'),
 defaultTag: npm.config.get('tag'),
 get tag () {
- log.warn('FIXME', 'using tag option, should be defaultTag')


nit: I thought we wanted this? At least until we have more stability in npm@7?

The problem is that it warns every time we do ...npm.flatOptions, which we do a lot of times. And we can't make the option non-enumerable, or else it won't be carried over in those cases. We'd have to do a custom clone function that does Object.getOwnPropertyDescriptors and turns the setter into a setter on the resulting object, and then use that everywhere, but then our deps also do ...opts all over the place, so actually making that change would be almost as bad as figgy-pudding.
It was a reasonable idea, but doesn't actually work, turns out.

Totally makes sense, I hadn't realised that. Completely agree we should just pull this out for now. Might be worth just commenting it, so we remember where our warnings were?

Sure, if you wanna, go for it. I was just gonna leave it as a pre-v7 todo item to remove the getters from here and grep for any usage of those field names.

isaacs · 2020-03-05T06:52:10Z

Fixed the nits, replaced the read-json util with parse-json-even-better-errors, found 2 more bluebird .nodeify() calls that I'd overlooked.

mikemimik

LGTM

mikemimik · 2020-03-05T18:35:20Z

- opts = TeamConfig(opts).concat({description: null})
+ // XXX: "description" option to libnpmteam is used as a description of the
+ // team, but in npm's options, this is a boolean meaning "show the
+ // description in npm search output". Hence its being set to null here.


Not important, but eventually we can remove this.

@isaacs

PR-URL: #968 Credit: @isaacs and @mikemimikClose: #968

@isaacs

PR-URL: #968 Credit: @isaacs and @mikemimikClose: #968

@isaacs

PR-URL: #968 Credit: @isaacs and @mikemimikClose: #968

isaacs · 2020-03-03T08:12:36Z

Refactor of mikemimik/update-libnpmhook branch from pairing earlier today. Removes a lot of figgy-pudding and blue birds.

Commands updated with a new lib:

hook
access
team
org
search

Other random stuff:

cache.js
deprecate.js
fund.js
logout.js
owner.js
ping.js
star.js
stars.js
version.js
view.js
whoami.js

Utils:

utils/otplease.js
utils/git.js
utils/pulse-till-done.js
utils/read-user-info.js

Still todo:

publish (libnpmpublish)
unpublish (also libnpmpublish)
pack (should leverage pacote more than it does)
audit (should use @npmcli/arborist)
shrinkwrap (should use @npmcli/arborist)
update (should use @npmcli/arborist)
outdated (should use @npmcli/arborist)
ci (should use @npmcli/arborist)

isaacs · 2020-03-03T17:14:34Z

Note: this closes #935 as well.

mikemimik

Some discussion points, @isaacs let me know if you want me to handle any of this minutia.

mikemimik · 2020-03-04T16:54:30Z

 const readUserInfo = require('./read-user-info.js')

+const isOtpError = err =>
+ err.code === 'EOTP' || err.code === 'E401' && /one-time pass/.test(err.body)


Silly question, does this need brackets to evaluate correctly?
Suggested change
err.code==='EOTP'||err.code==='E401'&&/one-timepass/.test(err.body)
err.code==='EOTP'||(err.code==='E401'&&/one-timepass/.test(err.body))

It does not. && has higher priority than ||. In most expressions of boolean calculus, you can think of and (or any intersection operator) like multiplication or division, and or (or any union operator) like addition or subtraction. So a || b && c || d is like a + b * c + d, so it's like a + (b * c) + d and thus a || (b && c) || d.

mikemimik · 2020-03-05T03:46:56Z

 strictSSL: npm.config.get('strict-ssl'),
 defaultTag: npm.config.get('tag'),
 get tag () {
- log.warn('FIXME', 'using tag option, should be defaultTag')


nit: I thought we wanted this? At least until we have more stability in npm@7?

The problem is that it warns every time we do ...npm.flatOptions, which we do a lot of times. And we can't make the option non-enumerable, or else it won't be carried over in those cases. We'd have to do a custom clone function that does Object.getOwnPropertyDescriptors and turns the setter into a setter on the resulting object, and then use that everywhere, but then our deps also do ...opts all over the place, so actually making that change would be almost as bad as figgy-pudding.
It was a reasonable idea, but doesn't actually work, turns out.

Totally makes sense, I hadn't realised that. Completely agree we should just pull this out for now. Might be worth just commenting it, so we remember where our warnings were?

Sure, if you wanna, go for it. I was just gonna leave it as a pre-v7 todo item to remove the getters from here and grep for any usage of those field names.

isaacs · 2020-03-05T06:52:10Z

Fixed the nits, replaced the read-json util with parse-json-even-better-errors, found 2 more bluebird .nodeify() calls that I'd overlooked.

mikemimik

LGTM

mikemimik · 2020-03-05T18:35:20Z

- opts = TeamConfig(opts).concat({description: null})
+ // XXX: "description" option to libnpmteam is used as a description of the
+ // team, but in npm's options, this is a boolean meaning "show the
+ // description in npm search output". Hence its being set to null here.


Not important, but eventually we can remove this.

@isaacs

PR-URL: #968 Credit: @isaacs and @mikemimikClose: #968

@isaacs

PR-URL: #968 Credit: @isaacs and @mikemimikClose: #968

@isaacs

PR-URL: #968 Credit: @isaacs and @mikemimikClose: #968

isaacs · 2020-03-03T08:12:36Z

Refactor of mikemimik/update-libnpmhook branch from pairing earlier today. Removes a lot of figgy-pudding and blue birds.

Commands updated with a new lib:

hook
access
team
org
search

Other random stuff:

cache.js
deprecate.js
fund.js
logout.js
owner.js
ping.js
star.js
stars.js
version.js
view.js
whoami.js

Utils:

utils/otplease.js
utils/git.js
utils/pulse-till-done.js
utils/read-user-info.js

Still todo:

publish (libnpmpublish)
unpublish (also libnpmpublish)
pack (should leverage pacote more than it does)
audit (should use @npmcli/arborist)
shrinkwrap (should use @npmcli/arborist)
update (should use @npmcli/arborist)
outdated (should use @npmcli/arborist)
ci (should use @npmcli/arborist)

isaacs · 2020-03-03T17:14:34Z

Note: this closes #935 as well.

mikemimik

Some discussion points, @isaacs let me know if you want me to handle any of this minutia.

mikemimik · 2020-03-04T16:54:30Z

 const readUserInfo = require('./read-user-info.js')

+const isOtpError = err =>
+ err.code === 'EOTP' || err.code === 'E401' && /one-time pass/.test(err.body)


Silly question, does this need brackets to evaluate correctly?
Suggested change
err.code==='EOTP'||err.code==='E401'&&/one-timepass/.test(err.body)
err.code==='EOTP'||(err.code==='E401'&&/one-timepass/.test(err.body))

It does not. && has higher priority than ||. In most expressions of boolean calculus, you can think of and (or any intersection operator) like multiplication or division, and or (or any union operator) like addition or subtraction. So a || b && c || d is like a + b * c + d, so it's like a + (b * c) + d and thus a || (b && c) || d.

mikemimik · 2020-03-05T03:46:56Z

 strictSSL: npm.config.get('strict-ssl'),
 defaultTag: npm.config.get('tag'),
 get tag () {
- log.warn('FIXME', 'using tag option, should be defaultTag')


nit: I thought we wanted this? At least until we have more stability in npm@7?

The problem is that it warns every time we do ...npm.flatOptions, which we do a lot of times. And we can't make the option non-enumerable, or else it won't be carried over in those cases. We'd have to do a custom clone function that does Object.getOwnPropertyDescriptors and turns the setter into a setter on the resulting object, and then use that everywhere, but then our deps also do ...opts all over the place, so actually making that change would be almost as bad as figgy-pudding.
It was a reasonable idea, but doesn't actually work, turns out.

Totally makes sense, I hadn't realised that. Completely agree we should just pull this out for now. Might be worth just commenting it, so we remember where our warnings were?

Sure, if you wanna, go for it. I was just gonna leave it as a pre-v7 todo item to remove the getters from here and grep for any usage of those field names.

isaacs · 2020-03-05T06:52:10Z

Fixed the nits, replaced the read-json util with parse-json-even-better-errors, found 2 more bluebird .nodeify() calls that I'd overlooked.

mikemimik

LGTM

mikemimik · 2020-03-05T18:35:20Z

- opts = TeamConfig(opts).concat({description: null})
+ // XXX: "description" option to libnpmteam is used as a description of the
+ // team, but in npm's options, this is a boolean meaning "show the
+ // description in npm search output". Hence its being set to null here.


Not important, but eventually we can remove this.

@isaacs

PR-URL: #968 Credit: @isaacs and @mikemimikClose: #968

@isaacs

PR-URL: #968 Credit: @isaacs and @mikemimikClose: #968

@isaacs

PR-URL: #968 Credit: @isaacs and @mikemimikClose: #968

isaacs · 2020-03-03T08:12:36Z

Refactor of mikemimik/update-libnpmhook branch from pairing earlier today. Removes a lot of figgy-pudding and blue birds.

Commands updated with a new lib:

hook
access
team
org
search

Other random stuff:

cache.js
deprecate.js
fund.js
logout.js
owner.js
ping.js
star.js
stars.js
version.js
view.js
whoami.js

Utils:

utils/otplease.js
utils/git.js
utils/pulse-till-done.js
utils/read-user-info.js

Still todo:

publish (libnpmpublish)
unpublish (also libnpmpublish)
pack (should leverage pacote more than it does)
audit (should use @npmcli/arborist)
shrinkwrap (should use @npmcli/arborist)
update (should use @npmcli/arborist)
outdated (should use @npmcli/arborist)
ci (should use @npmcli/arborist)

isaacs · 2020-03-03T17:14:34Z

Note: this closes #935 as well.

mikemimik

Some discussion points, @isaacs let me know if you want me to handle any of this minutia.

mikemimik · 2020-03-04T16:54:30Z

 const readUserInfo = require('./read-user-info.js')

+const isOtpError = err =>
+ err.code === 'EOTP' || err.code === 'E401' && /one-time pass/.test(err.body)


Silly question, does this need brackets to evaluate correctly?
Suggested change
err.code==='EOTP'||err.code==='E401'&&/one-timepass/.test(err.body)
err.code==='EOTP'||(err.code==='E401'&&/one-timepass/.test(err.body))

It does not. && has higher priority than ||. In most expressions of boolean calculus, you can think of and (or any intersection operator) like multiplication or division, and or (or any union operator) like addition or subtraction. So a || b && c || d is like a + b * c + d, so it's like a + (b * c) + d and thus a || (b && c) || d.

mikemimik · 2020-03-05T03:46:56Z

 strictSSL: npm.config.get('strict-ssl'),
 defaultTag: npm.config.get('tag'),
 get tag () {
- log.warn('FIXME', 'using tag option, should be defaultTag')


nit: I thought we wanted this? At least until we have more stability in npm@7?

The problem is that it warns every time we do ...npm.flatOptions, which we do a lot of times. And we can't make the option non-enumerable, or else it won't be carried over in those cases. We'd have to do a custom clone function that does Object.getOwnPropertyDescriptors and turns the setter into a setter on the resulting object, and then use that everywhere, but then our deps also do ...opts all over the place, so actually making that change would be almost as bad as figgy-pudding.
It was a reasonable idea, but doesn't actually work, turns out.

Totally makes sense, I hadn't realised that. Completely agree we should just pull this out for now. Might be worth just commenting it, so we remember where our warnings were?

Sure, if you wanna, go for it. I was just gonna leave it as a pre-v7 todo item to remove the getters from here and grep for any usage of those field names.

isaacs · 2020-03-05T06:52:10Z

Fixed the nits, replaced the read-json util with parse-json-even-better-errors, found 2 more bluebird .nodeify() calls that I'd overlooked.

mikemimik

LGTM

mikemimik · 2020-03-05T18:35:20Z

- opts = TeamConfig(opts).concat({description: null})
+ // XXX: "description" option to libnpmteam is used as a description of the
+ // team, but in npm's options, this is a boolean meaning "show the
+ // description in npm search output". Hence its being set to null here.


Not important, but eventually we can remove this.

@isaacs

PR-URL: #968 Credit: @isaacs and @mikemimikClose: #968

@isaacs

PR-URL: #968 Credit: @isaacs and @mikemimikClose: #968

@isaacs

PR-URL: #968 Credit: @isaacs and @mikemimikClose: #968

isaacs · 2020-03-03T08:12:36Z

Refactor of mikemimik/update-libnpmhook branch from pairing earlier today. Removes a lot of figgy-pudding and blue birds.

Commands updated with a new lib:

hook
access
team
org
search

Other random stuff:

cache.js
deprecate.js
fund.js
logout.js
owner.js
ping.js
star.js
stars.js
version.js
view.js
whoami.js

Utils:

utils/otplease.js
utils/git.js
utils/pulse-till-done.js
utils/read-user-info.js

Still todo:

publish (libnpmpublish)
unpublish (also libnpmpublish)
pack (should leverage pacote more than it does)
audit (should use @npmcli/arborist)
shrinkwrap (should use @npmcli/arborist)
update (should use @npmcli/arborist)
outdated (should use @npmcli/arborist)
ci (should use @npmcli/arborist)

isaacs · 2020-03-03T17:14:34Z

Note: this closes #935 as well.

mikemimik

Some discussion points, @isaacs let me know if you want me to handle any of this minutia.

mikemimik · 2020-03-04T16:54:30Z

 const readUserInfo = require('./read-user-info.js')

+const isOtpError = err =>
+ err.code === 'EOTP' || err.code === 'E401' && /one-time pass/.test(err.body)


Silly question, does this need brackets to evaluate correctly?
Suggested change
err.code==='EOTP'||err.code==='E401'&&/one-timepass/.test(err.body)
err.code==='EOTP'||(err.code==='E401'&&/one-timepass/.test(err.body))

It does not. && has higher priority than ||. In most expressions of boolean calculus, you can think of and (or any intersection operator) like multiplication or division, and or (or any union operator) like addition or subtraction. So a || b && c || d is like a + b * c + d, so it's like a + (b * c) + d and thus a || (b && c) || d.

mikemimik · 2020-03-05T03:46:56Z

 strictSSL: npm.config.get('strict-ssl'),
 defaultTag: npm.config.get('tag'),
 get tag () {
- log.warn('FIXME', 'using tag option, should be defaultTag')


nit: I thought we wanted this? At least until we have more stability in npm@7?

The problem is that it warns every time we do ...npm.flatOptions, which we do a lot of times. And we can't make the option non-enumerable, or else it won't be carried over in those cases. We'd have to do a custom clone function that does Object.getOwnPropertyDescriptors and turns the setter into a setter on the resulting object, and then use that everywhere, but then our deps also do ...opts all over the place, so actually making that change would be almost as bad as figgy-pudding.
It was a reasonable idea, but doesn't actually work, turns out.

Totally makes sense, I hadn't realised that. Completely agree we should just pull this out for now. Might be worth just commenting it, so we remember where our warnings were?

Sure, if you wanna, go for it. I was just gonna leave it as a pre-v7 todo item to remove the getters from here and grep for any usage of those field names.

isaacs · 2020-03-05T06:52:10Z

Fixed the nits, replaced the read-json util with parse-json-even-better-errors, found 2 more bluebird .nodeify() calls that I'd overlooked.

mikemimik

LGTM

mikemimik · 2020-03-05T18:35:20Z

- opts = TeamConfig(opts).concat({description: null})
+ // XXX: "description" option to libnpmteam is used as a description of the
+ // team, but in npm's options, this is a boolean meaning "show the
+ // description in npm search output". Hence its being set to null here.


Not important, but eventually we can remove this.

@isaacs

PR-URL: #968 Credit: @isaacs and @mikemimikClose: #968

@isaacs

PR-URL: #968 Credit: @isaacs and @mikemimikClose: #968

@isaacs

PR-URL: #968 Credit: @isaacs and @mikemimikClose: #968

isaacs · 2020-03-03T08:12:36Z

Refactor of mikemimik/update-libnpmhook branch from pairing earlier today. Removes a lot of figgy-pudding and blue birds.

Commands updated with a new lib:

hook
access
team
org
search

Other random stuff:

cache.js
deprecate.js
fund.js
logout.js
owner.js
ping.js
star.js
stars.js
version.js
view.js
whoami.js

Utils:

utils/otplease.js
utils/git.js
utils/pulse-till-done.js
utils/read-user-info.js

Still todo:

publish (libnpmpublish)
unpublish (also libnpmpublish)
pack (should leverage pacote more than it does)
audit (should use @npmcli/arborist)
shrinkwrap (should use @npmcli/arborist)
update (should use @npmcli/arborist)
outdated (should use @npmcli/arborist)
ci (should use @npmcli/arborist)

isaacs · 2020-03-03T17:14:34Z

Note: this closes #935 as well.

mikemimik

Some discussion points, @isaacs let me know if you want me to handle any of this minutia.

mikemimik · 2020-03-04T16:54:30Z

 const readUserInfo = require('./read-user-info.js')

+const isOtpError = err =>
+ err.code === 'EOTP' || err.code === 'E401' && /one-time pass/.test(err.body)


Silly question, does this need brackets to evaluate correctly?
Suggested change
err.code==='EOTP'||err.code==='E401'&&/one-timepass/.test(err.body)
err.code==='EOTP'||(err.code==='E401'&&/one-timepass/.test(err.body))

It does not. && has higher priority than ||. In most expressions of boolean calculus, you can think of and (or any intersection operator) like multiplication or division, and or (or any union operator) like addition or subtraction. So a || b && c || d is like a + b * c + d, so it's like a + (b * c) + d and thus a || (b && c) || d.

mikemimik · 2020-03-05T03:46:56Z

 strictSSL: npm.config.get('strict-ssl'),
 defaultTag: npm.config.get('tag'),
 get tag () {
- log.warn('FIXME', 'using tag option, should be defaultTag')


nit: I thought we wanted this? At least until we have more stability in npm@7?

The problem is that it warns every time we do ...npm.flatOptions, which we do a lot of times. And we can't make the option non-enumerable, or else it won't be carried over in those cases. We'd have to do a custom clone function that does Object.getOwnPropertyDescriptors and turns the setter into a setter on the resulting object, and then use that everywhere, but then our deps also do ...opts all over the place, so actually making that change would be almost as bad as figgy-pudding.
It was a reasonable idea, but doesn't actually work, turns out.

Totally makes sense, I hadn't realised that. Completely agree we should just pull this out for now. Might be worth just commenting it, so we remember where our warnings were?

Sure, if you wanna, go for it. I was just gonna leave it as a pre-v7 todo item to remove the getters from here and grep for any usage of those field names.

isaacs · 2020-03-05T06:52:10Z

Fixed the nits, replaced the read-json util with parse-json-even-better-errors, found 2 more bluebird .nodeify() calls that I'd overlooked.

mikemimik

LGTM

mikemimik · 2020-03-05T18:35:20Z

- opts = TeamConfig(opts).concat({description: null})
+ // XXX: "description" option to libnpmteam is used as a description of the
+ // team, but in npm's options, this is a boolean meaning "show the
+ // description in npm search output". Hence its being set to null here.


Not important, but eventually we can remove this.

@isaacs

PR-URL: #968 Credit: @isaacs and @mikemimikClose: #968

@isaacs

PR-URL: #968 Credit: @isaacs and @mikemimikClose: #968

@isaacs

PR-URL: #968 Credit: @isaacs and @mikemimikClose: #968

isaacs · 2020-03-03T08:12:36Z

Refactor of mikemimik/update-libnpmhook branch from pairing earlier today. Removes a lot of figgy-pudding and blue birds.

Commands updated with a new lib:

hook
access
team
org
search

Other random stuff:

cache.js
deprecate.js
fund.js
logout.js
owner.js
ping.js
star.js
stars.js
version.js
view.js
whoami.js

Utils:

utils/otplease.js
utils/git.js
utils/pulse-till-done.js
utils/read-user-info.js

Still todo:

publish (libnpmpublish)
unpublish (also libnpmpublish)
pack (should leverage pacote more than it does)
audit (should use @npmcli/arborist)
shrinkwrap (should use @npmcli/arborist)
update (should use @npmcli/arborist)
outdated (should use @npmcli/arborist)
ci (should use @npmcli/arborist)

isaacs · 2020-03-03T17:14:34Z

Note: this closes #935 as well.

mikemimik

Some discussion points, @isaacs let me know if you want me to handle any of this minutia.

mikemimik · 2020-03-04T16:54:30Z

 const readUserInfo = require('./read-user-info.js')

+const isOtpError = err =>
+ err.code === 'EOTP' || err.code === 'E401' && /one-time pass/.test(err.body)


Silly question, does this need brackets to evaluate correctly?
Suggested change
err.code==='EOTP'||err.code==='E401'&&/one-timepass/.test(err.body)
err.code==='EOTP'||(err.code==='E401'&&/one-timepass/.test(err.body))

It does not. && has higher priority than ||. In most expressions of boolean calculus, you can think of and (or any intersection operator) like multiplication or division, and or (or any union operator) like addition or subtraction. So a || b && c || d is like a + b * c + d, so it's like a + (b * c) + d and thus a || (b && c) || d.

mikemimik · 2020-03-05T03:46:56Z

 strictSSL: npm.config.get('strict-ssl'),
 defaultTag: npm.config.get('tag'),
 get tag () {
- log.warn('FIXME', 'using tag option, should be defaultTag')


nit: I thought we wanted this? At least until we have more stability in npm@7?

The problem is that it warns every time we do ...npm.flatOptions, which we do a lot of times. And we can't make the option non-enumerable, or else it won't be carried over in those cases. We'd have to do a custom clone function that does Object.getOwnPropertyDescriptors and turns the setter into a setter on the resulting object, and then use that everywhere, but then our deps also do ...opts all over the place, so actually making that change would be almost as bad as figgy-pudding.
It was a reasonable idea, but doesn't actually work, turns out.

Totally makes sense, I hadn't realised that. Completely agree we should just pull this out for now. Might be worth just commenting it, so we remember where our warnings were?

Sure, if you wanna, go for it. I was just gonna leave it as a pre-v7 todo item to remove the getters from here and grep for any usage of those field names.

isaacs · 2020-03-05T06:52:10Z

Fixed the nits, replaced the read-json util with parse-json-even-better-errors, found 2 more bluebird .nodeify() calls that I'd overlooked.

mikemimik

LGTM

mikemimik · 2020-03-05T18:35:20Z

- opts = TeamConfig(opts).concat({description: null})
+ // XXX: "description" option to libnpmteam is used as a description of the
+ // team, but in npm's options, this is a boolean meaning "show the
+ // description in npm search output". Hence its being set to null here.


Not important, but eventually we can remove this.

@isaacs

PR-URL: #968 Credit: @isaacs and @mikemimikClose: #968

@isaacs

PR-URL: #968 Credit: @isaacs and @mikemimikClose: #968

@isaacs

PR-URL: #968 Credit: @isaacs and @mikemimikClose: #968

isaacs · 2020-03-03T08:12:36Z

Refactor of mikemimik/update-libnpmhook branch from pairing earlier today. Removes a lot of figgy-pudding and blue birds.

Commands updated with a new lib:

hook
access
team
org
search

Other random stuff:

cache.js
deprecate.js
fund.js
logout.js
owner.js
ping.js
star.js
stars.js
version.js
view.js
whoami.js

Utils:

utils/otplease.js
utils/git.js
utils/pulse-till-done.js
utils/read-user-info.js

Still todo:

publish (libnpmpublish)
unpublish (also libnpmpublish)
pack (should leverage pacote more than it does)
audit (should use @npmcli/arborist)
shrinkwrap (should use @npmcli/arborist)
update (should use @npmcli/arborist)
outdated (should use @npmcli/arborist)
ci (should use @npmcli/arborist)

isaacs · 2020-03-03T17:14:34Z

Note: this closes #935 as well.

mikemimik

Some discussion points, @isaacs let me know if you want me to handle any of this minutia.

mikemimik · 2020-03-04T16:54:30Z

 const readUserInfo = require('./read-user-info.js')

+const isOtpError = err =>
+ err.code === 'EOTP' || err.code === 'E401' && /one-time pass/.test(err.body)


Silly question, does this need brackets to evaluate correctly?
Suggested change
err.code==='EOTP'||err.code==='E401'&&/one-timepass/.test(err.body)
err.code==='EOTP'||(err.code==='E401'&&/one-timepass/.test(err.body))

It does not. && has higher priority than ||. In most expressions of boolean calculus, you can think of and (or any intersection operator) like multiplication or division, and or (or any union operator) like addition or subtraction. So a || b && c || d is like a + b * c + d, so it's like a + (b * c) + d and thus a || (b && c) || d.

mikemimik · 2020-03-05T03:46:56Z

 strictSSL: npm.config.get('strict-ssl'),
 defaultTag: npm.config.get('tag'),
 get tag () {
- log.warn('FIXME', 'using tag option, should be defaultTag')


nit: I thought we wanted this? At least until we have more stability in npm@7?

The problem is that it warns every time we do ...npm.flatOptions, which we do a lot of times. And we can't make the option non-enumerable, or else it won't be carried over in those cases. We'd have to do a custom clone function that does Object.getOwnPropertyDescriptors and turns the setter into a setter on the resulting object, and then use that everywhere, but then our deps also do ...opts all over the place, so actually making that change would be almost as bad as figgy-pudding.
It was a reasonable idea, but doesn't actually work, turns out.

Totally makes sense, I hadn't realised that. Completely agree we should just pull this out for now. Might be worth just commenting it, so we remember where our warnings were?

Sure, if you wanna, go for it. I was just gonna leave it as a pre-v7 todo item to remove the getters from here and grep for any usage of those field names.

isaacs · 2020-03-05T06:52:10Z

Fixed the nits, replaced the read-json util with parse-json-even-better-errors, found 2 more bluebird .nodeify() calls that I'd overlooked.

mikemimik

LGTM

mikemimik · 2020-03-05T18:35:20Z

- opts = TeamConfig(opts).concat({description: null})
+ // XXX: "description" option to libnpmteam is used as a description of the
+ // team, but in npm's options, this is a boolean meaning "show the
+ // description in npm search output". Hence its being set to null here.


Not important, but eventually we can remove this.

@isaacs

PR-URL: #968 Credit: @isaacs and @mikemimikClose: #968

@isaacs

PR-URL: #968 Credit: @isaacs and @mikemimikClose: #968

@isaacs

PR-URL: #968 Credit: @isaacs and @mikemimikClose: #968

isaacs requested a review from a team as a code owner March 3, 2020 08:12

mikemimik added this to the OSS - Sprint 5 milestone Mar 3, 2020

mikemimik added Enhancement new feature or improvement Release 7.x work is associated with a specific npm 7 release semver:major backwards-incompatible breaking changes labels Mar 3, 2020

mikemimik linked an issue Mar 3, 2020 that may be closed by this pull request
Update libnpm* packages in the cli #970
Closed
6 tasks

mikemimik self-assigned this Mar 3, 2020

mikemimik force-pushed the isaacs/update-libnpmetc-stuff branch 3 times, most recently from c6d224f to 924d85bCompare March 5, 2020 03:52

mikemimik suggested changes Mar 5, 2020
View reviewed changes

isaacs commented Mar 5, 2020
View reviewed changes

Comment threadlib/ping.js

mikemimik approved these changes Mar 5, 2020
View reviewed changes

isaacs added a commit that referenced this pull request Mar 6, 2020
Remove two remaining BB remnants
0bac7ed
PR-URL: #968 Credit: @isaacs and @mikemimikClose: #968

isaacs added a commit that referenced this pull request Mar 7, 2020
Remove two remaining BB remnants
bf3370b
PR-URL: #968 Credit: @isaacs and @mikemimikClose: #968

isaacs closed this Mar 9, 2020

isaacs force-pushed the isaacs/update-libnpmetc-stuff branch from 53b67a1 to bf3370bCompare March 9, 2020 16:41

isaacs added a commit that referenced this pull request May 8, 2020
Remove two remaining BB remnants
7df93b7
PR-URL: #968 Credit: @isaacs and @mikemimikClose: #968

isaacs deleted the isaacs/update-libnpmetc-stuff branch October 2, 2020 21:43

snyk-ioBot mentioned this pull request Oct 23, 2024
[Snyk] Fix for 36 vulnerabilities Graybar-codespace/cli-3#153
Open