file are not sorted in ascending order for numeric file name

[rajeevbarde]

• VSCode Version: 1.47.3
• OS Version: Windows 10 v2004

On VS Explorer, the file is not sorted in ascending order as shown below, (compared with Window Explorer)

Does this issue occur when all extensions are disabled?: Yes

[vscodebot]

(Experimental duplicate detection)
Thanks for submitting this issue. Please also check if it is already covered by an existing one, like:

• editor.action.sortLinesAscending has confusing sort order for symbols (#48123)
• Sorting and Grouping files in Explorer (#100405)

[isidorn]

We are fine tuning our sorting.
I believe we have not covered this case. Thus assigning to Backlog and we might tackle in the future.

fyi @leilapearson

[leilapearson]

Thanks for the report @rajeevbarde

Yes, unfortunately numbers with a decimal separator will not always sort numerically. This is expected.

There are two things at play here:

1. We're using a standard javascript localeCompare() with numeric sorting enabled, which essentially works by ignoring leading zeros when comparing number segments. That means that 1 < 02 < 003 (good) and 1.1 < 1.02 < 1.003 (maybe not good)

2. We're disambiguating equivalent numbers (same number but different number of leading zeros) by length - and putting the shorter numbers first - so 1 < 01 < 001 and 1.1 < 1.01 < 1.001.

(1) is tricky. The reason I say "maybe not good" instead of just "not good" is that the decimal character and thousands separator characters differ by locale and also contextually. For example, a dot in a version number or in an IP address is not a decimal separator - it just separates number groups. I think this is why localeCompare() doesn't bother trying to account for decimal separators and thousands separators in different locales.

(2) would be easy enough to change. I could just reverse it so that longer equivalent numbers come first instead of last. I thought it was nicer to sort shorter equivalent numbers first - and more consistent with - for example - sorting a < aa < aaa. It is probably more common to sort them last though.

The simple change for (2) would resolve the specific cases in this report but if you expand the cases a bit you'll still see 2.01 < 2.1 < 2.02 < 2.2 - not the (possibly) desired 2.01 < 2.02 < 2.1 < 2.2 - so numbers with decimals would still not be in numerical order. Thus I'm not sure that a change would be worthwhile.

[ronakg]

@leilapearson I want to report another sorting issue here. Let me know if I should create a new bug for this.

For a case where 2 filenames have the same prefix and 1 of them doesn't have any suffix at all, the file without the suffix should be sorted ahead of the one with a suffix. A file named prefix.ext should be sorted ahead of prefixSuffix.ext or prefix_suffix.ext.

Take a look at the following example. close.go, close_test.go and close_channel.go are not sorted alphanumerically.

VSCode sorting -

Bash sorting -

[leilapearson]

Hi @ronakg. The reason this happens is that vscode sorts in locale order but bash shorts in unicode order. Locale order sorts the underscore character before the dot character, but unicode order does the opposite. I submitted a PR that adds the option for vscode to sort in unicode order - as well as options to sort in locale order but group by uppercase first or lowercase first. That PR is currently on hold though as the maintainers evaluate the level of demand for these new settings.

See PR #97272 for more detail. Issue #99955 also provides a lot more background if you're interested.

[ronakg]

@leilapearson that option would be nice. Thanks for the info.

[fastrde]

This is what happened for me with Sort Order = default
Looks like the ordering is by the value of the first digits and not alphanumerical.

9658.. < 09765.. < 9778... < 12705... and so on

Version: 1.52.1
Commit: ea3859d
Date: 2020-12-16T16:30:02.420Z (3 wks ago)
Electron: 9.3.5
Chrome: 83.0.4103.122
Node.js: 12.14.1
V8: 8.3.110.13-electron.0
OS: Darwin x64 19.6.0

[leilapearson]

@fastrde Hex numbers aren't going to sort so nicely unfortunately... The default sort will sort numerically - ignoring leading zeros - and considering the number to be complete once it reaches anything non-numeric... The PR I mentioned above would provide a unicode sorting option that might give you what you are looking for... not a nice hex sort, but at least a character by character comparison sort.

[fastrde]

@leilapearson I'm looking for what is described as the default sorting in VSCode. I don't want to sort hex numbers by value. I don't want to sort numbers by value either. I want to sort alphabetical. That's all :-). The description of sort order default is Files and folders are sorted by their names, in alphabetical order. Folders are displayed before files.
So I hope, that the mentioned PRs implements the default behavior. At the moment the description didn't match with the applied sorting..

default sorting at the moment

01apple    (01apple   = numerical value  1...)
11coconut  (11coconut = numerical value 11...)
012banana  (012banana = numerical value 12...)

Sorting I expected when I read the description

012banana  (012banana = alphabetic 012...)
01apple    (01apple   = alphabetic 01a...)
11coconut  (11coconut = alphabetic 11c...)

[leilapearson]

@fastrde that makes sense. Unfortunately the text is misleading. The actual behavior - since way before I got involved - has been to do a localeCompare with the numeric option enabled. So no, it isn't a truly an alphabetical search. I guess someone could fix the text to better describe the actual behavior since it is somewhat misleading.

I did submit an earlier pull request that provided the ability to do the various sort options with the numeric option turned off, but it didn't seem like many people were asking for that option and the maintainers requested a less complicated PR so I withdrew it.

If enough people want a locale-based alphabetical comparison that doesn't have numeric sorting enabled - and the maintainers agree - then that feature could certainly be added at a later date.

[leilapearson]

Also, I'm happy to submit a new PR to add that new "non-numeric" option if the maintainers give the go ahead.

[paolieri]

A "non-numeric" option would be very useful; I was bitten by this too, in the same scenario as @fastrde (a directory with many files containing hex hashes in their name).

[paolieri]

@leilapearson was waiting for the go-ahead to submit a PR adding a "non-numeric" file sort option.
This would have the benefits of:

• preserving the current/default behavior (to avoid surprises)
• giving an option for those who need files in true alphabetical order (as ls would show them)

@isidorn I hope maintainers can reconsider this, I believe it is in the scope of VSCode

[isidorn]

I am discussing a potential PR with @leilapearson on another issue, once we tackle that we might look into this.

[Leedehai]

Hi @isidorn :) Wondering if there is a decision on the PR you mentioned above? Could you re-open this issue?

[leilapearson]

@Leedehai the PR he was referring to was merged earlier this month and is currently available in VS Code Insiders builds and should be in the next VS Code release which I think is due out pretty soon. It doesn't directly address this issue, but there is a unicode option which is a non-numeric sort which will perhaps meet your needs? If you have a chance to try it, let us know what you think.