Does ISO8601 require a separator in the time if one is used in the date?

1nickt has asked for the wisdom of the Perl Monks concerning the following question:

Hi all,

The doc for Regexp::Common::time ISO8601 datetimes states:

The date separator character may be a hyphen, slash (/), period, or empty string (omitted). The two date separators must match.

The time separator character may be a colon, a period, a space, or empty string (omitted). The two time separators must match.

However the standard for an ISO8601 datetime appears to additionally require that if a separator is used for the date or the time, a separator must also be used for the other component. I cannot find a spec that states this, but read on ...

Demonstrated with DateTime::Format::ISO8601 :

$ perl -Mstrict -MRegexp::Common=time -MDateTime::Format::ISO8601 -wE 
+'
my $str = "2018-03-01T110601Z";
say "DateTime::Format: ", eval { my $dt = DateTime::Format::ISO8601->p
+arse_datetime($str) } ? "valid" : "not valid: $@";
say "Regexp::Common: ", $str =~ /$RE{time}{iso}/ ? "valid" : "not vali
+d";
'

DateTime::Format: not valid: Invalid date format: 2018-03-01T110601Z a
+t -e line 3.
    eval {...} called at -e line 3

Regexp::Common: valid
[download]

Control:

$ perl -Mstrict -MRegexp::Common=time -MDateTime::Format::ISO8601 -wE 
+'
my $str = "2018-03-01T11:06:01Z";
say "DateTime::Format: ", eval { my $dt = DateTime::Format::ISO8601->p
+arse_datetime($str) } ? "valid" : "not valid: $@";
say "Regexp::Common: ", $str =~ /$RE{time}{iso}/ ? "valid" : "not vali
+d";
'

DateTime::Format: valid
Regexp::Common: valid
[download]

Which is right?

The way forward always starts with a minimal test.

Comment on Does ISO8601 require a separator in the time if one is used in the date? Select or Download Code

Replies are listed 'Best First'.
Re: Does ISO8601 require a separator in the time if one is used in the date? by hippo (Archbishop) on Mar 02, 2018 at 09:42 UTC
The discussion to this point appears to be heading in a direction of which our brothers over at lawmonks.org would be proud. In preference, why not just adhere to the Robustness Principle and use DateTime::Format::ISO8601 for your output but Regexp::Common for your input?	[reply]
Re^2: Does ISO8601 require a separator in the time if one is used in the date? by 1nickt (Canon) on Mar 02, 2018 at 11:40 UTC
Thanks hippo, it's a good point and in general a good rule. The problem is that if we accept scruffy datetimes with Regexp::Common::time then we can't parse them to prep for DB insertion using DateTime::Format::ISO8601 and the parsing becomes more complicated than it should be. Maybe the cleanest way would be to write a class that uses both of those modules and provides a single interface. The way forward always starts with a minimal test.	[reply]
Re^3: Does ISO8601 require a separator in the time if one is used in the date? by Anonymous Monk on Mar 02, 2018 at 12:06 UTC
The FAQ is o-on :) datetime natural	[reply]
Re: Does ISO8601 require a separator in the time if one is used in the date? by Eily (Monsignor) on Mar 01, 2018 at 16:46 UTC
From Wikipedia: Either basic or extended formats may be used, but both date and time must use the same format. (Basic format meaning without separators). There's no citation to go with it though.	[reply]
Re^2: Does ISO8601 require a separator in the time if one is used in the date? by ikegami (Patriarch) on Mar 01, 2018 at 18:29 UTC
On the other hand, DateTime::Format::ISO8601 does have a citation: Section 5.4 of ISO 8601:2000(E).	[reply]
Re^3: Does ISO8601 require a separator in the time if one is used in the date? by 1nickt (Canon) on Mar 01, 2018 at 20:13 UTC
Yes, I've seen that. Do you have a link to the spec (see original question)? The 2000 spec was superceded by the 2004 spec which is available for purchase at https://www.iso.org/obp/ui/#iso:std:iso:8601:ed-3:v1:en. But I can not find a published version of it. Edit: An older working doc: http://www.loc.gov/standards/datetime/iso-tc154-wg5_n0038_iso_wd_8601-1_2016-02-16.pdf. In this doc the only examples of "expanded" format datetimes have separators in both date and time. Another piece of circumstantial evidence ... The way forward always starts with a minimal test.	[reply]
Re^4: Does ISO8601 require a separator in the time if one is used in the date? by Eily (Monsignor) on Mar 02, 2018 at 09:16 UTC
Re^2: Does ISO8601 require a separator in the time if one is used in the date? by 1nickt (Canon) on Mar 01, 2018 at 16:55 UTC
Thanks Eily. I had seen that, but frankly I'd rank DateTime modules above Wikipedia as an authority. I'm hoping someone can point to the spec, although I've looked at that in various places as well. The way forward always starts with a minimal test.	[reply]
Re^3: Does ISO8601 require a separator in the time if one is used in the date? by Eily (Monsignor) on Mar 01, 2018 at 17:56 UTC
I'd rank DateTime modules above Wikipedia as an authority Well in this case they agree anyway :) . But I guess nothing beats a direct citation from the spec.	[reply]
Re^3: Does ISO8601 require a separator in the time if one is used in the date? by stevieb (Canon) on Mar 01, 2018 at 16:58 UTC
I can't even find a single RFC that refers to this forced condition. Now I've started, I'm curious to find something myself here if possible...	[reply]
Re^4: Does ISO8601 require a separator in the time if one is used in the date? by ikegami (Patriarch) on Mar 01, 2018 at 18:34 UTC