davidelang
commented
4 years ago Thank you for this work, I was fighting this just the other day and thinking that I needed to dig into it again.
rather than using extradata, I would suggest using the new json support in the v2 syntax
so instead of %key:name-value-list:|% do %key:name-value-list:{"separator":"|"}%
that way we can add 'assignment' (instead of it always being '=' allow for lists like foo:bar,baz:abc)
also, can this support a quoted string as the value? This would possibly be a third parameter "valuetype"
I can also see us adding 'ignore-whitespace' as an optional parameter.
these things don't need to be added now, but rather than using 'extradata', I'd like us to use a named field to support adding the others later without the separator being such an oddball.
if this only accepts a single character separator, it should complain if you feed it more. what would it take to support longer separators?
David Lang
On Tue, 17 Mar 2020, frikilax wrote:
Date: Tue, 17 Mar 2020 03:00:59 -0700 From: frikilax notifications@github.com Reply-To: rsyslog/liblognorm reply@reply.github.com To: rsyslog/liblognorm liblognorm@noreply.github.com Cc: Subscribed subscribed@noreply.github.com Subject: [rsyslog/liblognorm] [IMPROVEMENT] add separator character option to 'name-value-list' parser (#335)
Description
An extradata parameter is now available for the experimental perser name-value-list, this parameter allows to set a different character as separator between key/value couples.
What's new
Before, this parser could only handle logs like
a=1 b=2 c=text_without_spacesBut with this parameter, a character can be specified to act as the matching separator between key/value couples: rulesbase configuration:
rule=:%keyval:name-value-list:|%or
rule=:%{"name": "keyval", "type": "name-value-list", "extradata": "|"}%the sample:
a=1|b=2|c=text_without_spaces|d=text with spacesDetails
This change only takes the first character of your extradata field, so
rule=:%keyval:name-value-list:test%would not throw an error, but would split only on "t"
Credits
thanks to @KGuillemot for the help You can view, comment on, or merge this pull request online at:
https://github.com/rsyslog/liblognorm/pull/335
-- Commit Summary --
- name-value-list type:: add optional separator to override default space separation between key/value instances
- FIX:: compilation errors
-- File Changes --
M src/parser.c (46) M src/parser.h (2) M src/pdag.c (2)
-- Patch Links --
https://github.com/rsyslog/liblognorm/pull/335.patch https://github.com/rsyslog/liblognorm/pull/335.diff
frikilax
bdolez
rgerhards
Description
An extradata parameter is now available for the experimental parser name-value-list, this parameter allows to set a different character as separator between key/value couples.
What's new
Before, this parser could only handle logs like
a=1 b=2 c=text_without_spacesBut with this update, a parameter 'separator' can be specified to act as the matching separator between key/value couples, and a parameter 'assignator' as a separator between key and value.
for example:
or
are equivalent (by default, legacy 'extradata' is assigned to the 'separator' parameter in the first case) the matching samples will then be:
a=1|b=2|c=text_without_spaces|d=text with spaceskey=42|other_key=hello there|yet-another.key=i'm a value|When 'assignator' is left by default, it will match '=' and will check for valid naming of keys (allowing only alphanumeric characters, '.', '-' and '_'). In the case someone wants to match another character by setting 'assignator', the parser will loosen the restriction and only cut keys and values by separating between 'assignator' and 'separator'. However, key name validation will NOT occur !
Rulebase with 'assignator' defined:
examples of matching sample:
a=1|key_two=hello world|thirdkey:=a valid value containing an \| escaped pipe||fourth#key=42and result:One can note character escaping will still be respected.
Credits
thanks to @KGuillemot for the help thanks to @bdolez for the original work on quoting (see #234)