console: add table method

[devsnek]

Refs https://console.spec.whatwg.org/#table

new super fine console.table method

i've been using firefox as my model for how the outputs should look

some column ordering stuff:

• indexed properties before named properties unless properties is passed
• Values column is always last, independent from properties
• Key column always directly before Values column, independent from properties
• (index) and (iteration index) columns are always first, independent from properties

Checklist

• make -j4 test (UNIX), or vcbuild test (Windows) passes
• tests and/or benchmarks are included
• documentation is changed or added
• commit message follows commit guidelines

Affected core subsystem(s)

console

[Trott]

Thanks for doing this. Can you update the documentation for console.table() in console.md?

[TimothyGu]

function works just fine in Chrome: console.table([function a() {}, function b() {}]).

[TimothyGu]

Why are global and process singled out?

[devsnek]

they are objectively easier to read when not run through the table method, and the spec says Fall back to just logging the argument if it can’t be parsed as tabular. which i think is fairly true for both of those.

[BridgeAR]

I agree in this case that it is going to be easier to read. The question is if we want to have a special handling here. And there could theoretically be other things that are difficult to read as well because they have lots of properties.

In chrome window works as tableable and I think that is comparable. I guess most people will not add process or global so I would say either we should not special handle them at all or we add a limit that by checking the property number.

[TimothyGu]

This will exhaust O, preventing it from further usage; use the previewMapIterator and previewSetIterator from internal/v8 instead. Make sure to set a limit in how many elements to allow. See how it's used in util.format (the limit is 100 in this case):

node/lib/util.js

Lines 821 to 840 in c84582c

	functionformatCollectionIterator(preview,ctx,value,recurseTimes,
	visibleKeys,keys){
	varnextRecurseTimes=recurseTimes===null ? null : recurseTimes-1;
	varvals=preview(value,100);
	varoutput=[];
	for(constoofvals){
	output.push(formatValue(ctx,o,nextRecurseTimes));
	}
	returnoutput;
	}
	functionformatMapIterator(ctx,value,recurseTimes,visibleKeys,keys){
	returnformatCollectionIterator(previewMapIterator,ctx,value,recurseTimes,
	visibleKeys,keys);
	}
	functionformatSetIterator(ctx,value,recurseTimes,visibleKeys,keys){
	returnformatCollectionIterator(previewSetIterator,ctx,value,recurseTimes,
	visibleKeys,keys);
	}

[TimothyGu]

Did you mean return [...O.keys()];?

[TimothyGu]

Same as SetIterator

[TimothyGu]

Move to the top

[TimothyGu]

Needs more tests:

console.table([3,4,{a: 3}]);console.table([3,4,{a: 3}],['nonexistent']);console.table([3,4,{a: 3}],['Value']);console.table([3,4,{Value: 3}],['Value']);console.table([,,,,,,,,,4]);console.table([[,,,,,,,,,4]]);console.table([[,,,,,,,,,{}]]);// etc.// Check browsers for sensible behaviors.

Also needed to test are Map objects, Map Iterator objects, Set objects, Set Iterator objects.

[BridgeAR]

Ping @devsnek

[devsnek]

yeah sorry i've been meaning to get to this, its on my list

[devsnek]

@Trott@TimothyGu changes have been made

requested changes have been made

[BridgeAR]

I guess console.log should actually be console.table.

[BridgeAR]

Shall we maybe call it untableable?

[devsnek]

doesn't really make sense, the method below is already named that

[BridgeAR]

You misunderstood my comment: I did not mean the whole variable name but only the beginning. And that for all of these variables.

The difference is the e.

[devsnek]

oh. i have no idea. pretty sure tableable/tablable isn't english anyway so i have no idea where to go with that. i think following other words it would become tablible but that looks really wrong.

[BridgeAR]

Maybe @Trott has a opinion about it?

[BridgeAR]

I agree in this case that it is going to be easier to read. The question is if we want to have a special handling here. And there could theoretically be other things that are difficult to read as well because they have lots of properties.

In chrome window works as tableable and I think that is comparable. I guess most people will not add process or global so I would say either we should not special handle them at all or we add a limit that by checking the property number.

[BridgeAR]

The spec says it should log the properties. We could show those and add a special handling for them but e.g. Chrome does not show them.

The same applies to the entries below.

[BridgeAR]

This will always print \tundefined for Map and Set but it is possible to get the value from those.

[BridgeAR]

This can probably never be reached. The same applies to isSet.

[BridgeAR]

Chrome shows this as the following that I think would be great to have:

(index) Value a
0 3
1 4
2 3

[devsnek]

i don't understand the output of that table

[BridgeAR]

There is no undefined value for a in index 0 and 1. The value does not exist. But there is a value and we can show that with 3 and 4.

Just enter console.table([3, 4, { a: 3 }]) in chrome to see the output :-)

[devsnek]

rn the only disparity seems to be that i choose to ignore filters of props that don't exist while the browsers use undefined. i can change that.

[BridgeAR]

This is how it looks like for me:

[TimothyGu]

@BridgeAR There isn’t really a standard on what should be printed. As long as the output makes sense, which IMO it does, it’s good.

[BridgeAR]

@TimothyGu I am aware that there is no standard for it but I feel like the output used in chrome does make more sense and provides a relatively good user experience.

[apapirovski]

I'm in favour of the change requested by @BridgeAR here, in relation to missing vs undefined values. I think it would go a long way towards making this feature more useful.

[BridgeAR]

I think it would be best to add extra padding up to the value length that has the highest length of the column. That way it will improve the readability massively.

[devsnek]

i was considering that but i decided that as this isn't really intended to do more than humor the spec, and in general tables in terminals have a history of being formatted with single tabs without compensation (especially in configs and stuff) so i wanted to stay consistent with that.

[BridgeAR]

I feel like it improves the readability a lot, so I would really like to add that. I do not see any issues about inconsistencies here.

[BridgeAR]

Showing the value undefined is actually wrong here out of my perspective because there is just no value. So it should be empty instead of showing undefined. Otherwise this can not be distinguished from the value undefined.

As an extra test I would add:

test([{a: 1},{b: undefined}],'(index) a b'+'0 1 '+'1 undefined')

[BridgeAR]

I feel like we should handle iterators as not tableable.

[BridgeAR]

Ping @devsnek

[BridgeAR]

The booleans, undefined and NaN are actually redundant to the untablableTypes. So it would be best to remove the untablableValues and just explicitely test for null in the untableable function.

[BridgeAR]

Maybe @Trott has a opinion about it?

[BridgeAR]

This check has to be last. Otherwise it will fail for null and undefined. Please add a test for this as well.

[BridgeAR]

data[row] should be replaced with item here.

[BridgeAR]

Showing the value undefined is actually wrong here out of my perspective because there is just no value. So it should be empty instead of showing undefined. Otherwise this can not be distinguished from the value undefined.

As an extra test I would add:

test([{a: 1},{b: undefined}],'(index) a b'+'0 1 '+'1 undefined')

[devsnek]

@BridgeAR new code is here, i'm planning to clean up cli_table.js a bit but the behavior is where i want it to be

[BridgeAR]

undefined is redundant to value == null. Either the check above should be strict equal or the undefined should be removed here.

[TimothyGu]

string[]

[TimothyGu]

nit: Keys

[BridgeAR]

Actually: we might want to change Values to Value instead. Because it is a single value at the specific index position.

[BridgeAR]

I do not want to block this any further from landing. So LGTM. I would like to get some further adjustments in but I can open a PR for that at some point in the future.

[devsnek]

CI https://ci.nodejs.org/job/node-test-pull-request/13937/

[devsnek]

CI https://ci.nodejs.org/job/node-test-pull-request/13947/

[devsnek]

as a heads-up i will be landing this in a few hours

[devsnek]

landed in 97ace04

[targos]

Don't forget the semver-minor label for PRs that add new features ;)

[targos]

Should this be backported to v9.x-staging? If yes please follow the guide and raise a backport PR, if not let me know or add the dont-land-on label.

[vitaly-t]

Method console.table has been in Node.js for a very long time, but it was doing nothing. And now that there is an implementation, it raises a question of how to do a proper check that the current version of Node.js actually implements console.table?

[BridgeAR]

@vitaly-t I can think of two ways: 1) check for the version. Everything from 10.x on does something. It will likely be backported as well. So that might not be a reliable check. 2) Hook into stdout while a module is loaded and do a feature check.