Re: nnimap-split-fancy

Re: nnimap-split-fancy

[Adam Sjøgren]

On Thu, 16 Dec 2010 19:10:49 -0600, Tommy wrote:

> 		   ("from" ".*domain1\\.com" "emails-from-domain1)
                                                                 ^
                                                             Missing ".

> 		   ("from" ".*domain2\\.com" "emails-from-domain2)
                                                                 ^
                                                              Ditto.

I often find that the easiest way to debug regexp-matching in Emacs is
to start with something extremely simple that works, and then gradually
adding things I think should work until they don't - it helps
pinpointing where I have forgotten some \'s or something (I am used to
Perls regexp syntax, so remembering all the leaning toothpicks in Emacs
regexps is sometimes a problem :-))


  Best regards,

    Adam

-- 
 "Accept the mystery!"                                        Adam Sjøgren
                                                         asjo@koldfront.dk

Re: nnimap-split-fancy

[Richard Riley]

asjo@koldfront.dk (Adam Sjøgren) writes:

> On Thu, 16 Dec 2010 19:10:49 -0600, Tommy wrote:
>
>> 		   ("from" ".*domain1\\.com" "emails-from-domain1)
>                                                                  ^
>                                                              Missing ".
>
>> 		   ("from" ".*domain2\\.com" "emails-from-domain2)
>                                                                  ^
>                                                               Ditto.
>
> I often find that the easiest way to debug regexp-matching in Emacs is
> to start with something extremely simple that works, and then gradually
> adding things I think should work until they don't - it helps
> pinpointing where I have forgotten some \'s or something (I am used to
> Perls regexp syntax, so remembering all the leaning toothpicks in Emacs
> regexps is sometimes a problem :-))
>
>   Best regards,
>
>     Adam

There is also the wonderful regexp-builder.

regards

r.


-- 
☘ http://www.shamrockirishbar.com, http://splash-of-open-sauce.blogspot.com/ http://www.richardriley.net

Re: nnimap-split-fancy

[Tommy Kelly]

asjo@koldfront.dk (Adam Sjøgren) writes:

> On Thu, 16 Dec 2010 19:10:49 -0600, Tommy wrote:
>
>> 		   ("from" ".*domain1\\.com" "emails-from-domain1)
>                                                                  ^
>                                                              Missing ".

Thanks Adam, but that's my bad. I omitted the closing quotes as I
created my email message. The original .gnus was correct (font and
indentation would have shown up that error a mile away).

But it was, as I say, my bad. Ignore me for now while I prepare a
definitive problem case.

But a question in the meantime:

> ... it helps
> pinpointing where I have forgotten some \'s or something (I am used to
> Perls regexp syntax, so remembering all the leaning toothpicks in Emacs
> regexps is sometimes a problem :-))

Yeah, me too. For example: why does the . in domain1.com require the
"\\" when included in the fancy split approach, but no toothpicks at all
are needed when the same string is included in the normal (non fancy)
approach?

At first I assumed it was because in the fancy approach the string is a regexp
while in non-fancy it's just a plain string. But that's not true is it? The
use of things like "^" and ".*"  in non-fancy shows it's a regexp too,
no?

Tommy

About toothpicks (was Re: nnimap-split-fancy)

[Štěpán Němec]

Tommy Kelly <tommy.kelly@verilab.com> writes:

> asjo@koldfront.dk (Adam Sjøgren) writes:

>> ... it helps
>> pinpointing where I have forgotten some \'s or something (I am used to
>> Perls regexp syntax, so remembering all the leaning toothpicks in Emacs
>> regexps is sometimes a problem :-))
>
> Yeah, me too. For example: why does the . in domain1.com require the
> "\\" when included in the fancy split approach, but no toothpicks at all
> are needed when the same string is included in the normal (non fancy)
> approach?
>
> At first I assumed it was because in the fancy approach the string is a regexp
> while in non-fancy it's just a plain string. But that's not true is it? The
> use of things like "^" and ".*"  in non-fancy shows it's a regexp too,
> no?

It's really simple. '.' matches anything except newline, including a
dot. If you want to match a literal dot, you need '\.', but inside lisp
strings, double quotes and backslashes must be escaped with a backslash,
so what you end up with is "\\.".

  Štěpán

Re: About toothpicks (was Re: nnimap-split-fancy)

[Tommy Kelly]

Štěpán Němec <stepnem@gmail.com> writes:
> ...If you want to match a literal dot, you need '\.', but inside lisp
> strings, double quotes and backslashes must be escaped with a backslash,
> so what you end up with is "\\.".

Thanks Štěpán. I think I'd figure that but what was confusing me was the
fact that an unescaped dot was shown in examples for the non-fancy
split method. For example, in info "6.3.3 Splitting Mail" you see an
example of:

     ("list.\\1" "From:.* \\(.*\\)-list@majordomo.com")

So shouldn't the dot after "majordomo" be double escaped?

Tommy

Error in (info "(gnus)Splitting Mail")

[Štěpán Němec]

Tommy Kelly <tommy.kelly@verilab.com> writes:

> Thanks Štěpán. I think I'd figure that but what was confusing me was the
> fact that an unescaped dot was shown in examples for the non-fancy
> split method. For example, in info "6.3.3 Splitting Mail" you see an
> example of:
>
>      ("list.\\1" "From:.* \\(.*\\)-list@majordomo.com")
>
> So shouldn't the dot after "majordomo" be double escaped?

Yeah, that's obviously just an error in the documentation, but a mostly
harmless one, as "." matches the dot anyway...

Cc-ing the ding list.

  Štěpán

Re: Error in (info "(gnus)Splitting Mail")

[Lars Magne Ingebrigtsen]

Štěpán Němec <stepnem@gmail.com> writes:

>>      ("list.\\1" "From:.* \\(.*\\)-list@majordomo.com")
>>
>> So shouldn't the dot after "majordomo" be double escaped?
>
> Yeah, that's obviously just an error in the documentation, but a mostly
> harmless one, as "." matches the dot anyway...

It's more readable without the \\, and the chance for false positives is
minuscule, so I think I'll just leave it as is.

-- 
(domestic pets only, the antidote for overdose, milk.)
  larsi@gnus.org * Lars Magne Ingebrigtsen

Re: Error in (info "(gnus)Splitting Mail")

[Tommy Kelly]

Lars Magne Ingebrigtsen <larsi@gnus.org> writes:

> Štěpán Němec <stepnem@gmail.com> writes:
>
>>>      ("list.\\1" "From:.* \\(.*\\)-list@majordomo.com")
>>>
>>> So shouldn't the dot after "majordomo" be double escaped?
>>
>> Yeah, that's obviously just an error in the documentation, but a mostly
>> harmless one, as "." matches the dot anyway...
>
> It's more readable without the \\, and the chance for false positives is
> minuscule, so I think I'll just leave it as is.

So: 

The person who wrote the stuff, understands it in intense
detail without the documentation, and therefore has little or no need
of the documentation thinks the documentation is more readable as it is.

The person who is trying to learn the stuff, and must rely heavily on
said documentation thinks the documentation would be more readable with a
a minor fix.

Hmmm.

:-)

Tommy

Re: Error in (info "(gnus)Splitting Mail")

[Ted Zlatanov]

On Sat, 18 Dec 2010 16:33:47 -0600 Tommy Kelly <tommy.kelly@verilab.com> wrote: 

TK> The person who wrote the stuff, understands it in intense
TK> detail without the documentation, and therefore has little or no need
TK> of the documentation thinks the documentation is more readable as it is.

TK> The person who is trying to learn the stuff, and must rely heavily on
TK> said documentation thinks the documentation would be more readable with a
TK> a minor fix.

Štěpán was noting a minor issue: if mail comes from majordomoXcom (any
single character) instead of majordomo.com, it will match that splitting
rule.  I don't think Štěpán suggested it would make the documentation
more readable to fix that.

Ted

Re: Error in (info "(gnus)Splitting Mail")

[Štěpán Němec]

Ted Zlatanov <tzz@lifelogs.com> writes:

> On Sat, 18 Dec 2010 16:33:47 -0600 Tommy Kelly <tommy.kelly@verilab.com> wrote: 
>
> TK> The person who wrote the stuff, understands it in intense
> TK> detail without the documentation, and therefore has little or no need
> TK> of the documentation thinks the documentation is more readable as it is.
>
> TK> The person who is trying to learn the stuff, and must rely heavily on
> TK> said documentation thinks the documentation would be more readable with a
> TK> a minor fix.
>
> Štěpán was noting a minor issue: if mail comes from majordomoXcom (any
> single character) instead of majordomo.com, it will match that splitting
> rule.  I don't think Štěpán suggested it would make the documentation
> more readable to fix that.

I think Tommy's and Lars's notion of "readable" is slightly incompatible
here. In my opinion the error (yes, IMO it _is_ an error) should be
fixed at least for the reason this whole thread started -- it induced
major confusion for Tommy, who, upon seeing the unescaped dot, began to
wonder if there was some magic going on with the fancy splitting rules.

(Yes, it's funny to have discussions about things like this, but not
having sloppy examples in the docs is precisely the way to prevent such
discussions.)

  Štěpán

Re: Error in (info "(gnus)Splitting Mail")

[Tommy Kelly]

Ted Zlatanov <tzz@lifelogs.com> writes:

> On Sat, 18 Dec 2010 16:33:47 -0600 Tommy Kelly <tommy.kelly@verilab.com> wrote: 
>
> TK> The person who wrote the stuff, understands it in intense
> TK> detail without the documentation, and therefore has little or no need
> TK> of the documentation thinks the documentation is more readable as it is.
>
> TK> The person who is trying to learn the stuff, and must rely heavily on
> TK> said documentation thinks the documentation would be more readable with a
> TK> a minor fix.
>
> Štěpán was noting a minor issue: if mail comes from majordomoXcom (any
> single character) instead of majordomo.com, it will match that splitting
> rule.  I don't think Štěpán suggested it would make the documentation
> more readable to fix that.

You're right. Štěpán explained the thing to me and then flagged it to
The Documenters. *I* am suggesting that Štěpán's flagging was useful and
that, contrary to Lars's view, the proposed fix is worthwhile and makes
things more readable[1]. 

And my underlying point was that while Lars is an undoubted expert in
Gnus, *I* am an expert in being a new and clueless user of it.  The
documentation has an intended core audience, and it ain't Lars.

So Štěpán's point was one of regexp structure; mine, one of pedagogy.

All that said. I'm new here and have invested a mere fraction of the
hours in this project that y'all have. So don't hesitate to pat me on
the head and say "Shhhh! The grown-ups are speaking." :-)

Tommy

[1] Or rather, it makes it more consistent and understandable --
a.k.a. better. I'm with everyone else in that adding toothpicks rarely
makes for improved *read*-ability. But that -- toothpicks (and (to
further make the (undoubtedly made (albeit obliquely) by now) point)
parens) -- is the life we have chosen for ourselves.

Re: Error in (info "(gnus)Splitting Mail")

[Ted Zlatanov]

On Sun, 19 Dec 2010 09:53:14 -0600 Tommy Kelly <tommy.kelly@verilab.com> wrote: 

TK> All that said. I'm new here and have invested a mere fraction of the
TK> hours in this project that y'all have. So don't hesitate to pat me on
TK> the head and say "Shhhh! The grown-ups are speaking." :-)

Nah, that would be condescending.  We do passive-aggressive :)

Seriously, your feedback is appreciated.  You have invested a lot of
time fighting features and jargon that are implied or learned for many
of the Gnus regulars, so your experience helps us.  If we don't respond
or fix things quickly (speaking for myself) it's for lack of time, not
desire to help.

Ted

Re: Error in (info "(gnus)Splitting Mail")

[Steinar Bang]

>>>>> Ted Zlatanov <tzz@lifelogs.com>:

> ...  If we don't respond or fix things quickly (speaking for myself)
> it's for lack of time, not desire to help.

Well, in my case the reason for lack of usefulness in my responses, have
been lack of knowledge.  I stopped using fancy split in 2001 or
thereabouts. 

Re: Error in (info "(gnus)Splitting Mail")

[Richard Riley]

Ted Zlatanov <tzz@lifelogs.com> writes:

> On Sun, 19 Dec 2010 09:53:14 -0600 Tommy Kelly <tommy.kelly@verilab.com> wrote: 
>
> TK> All that said. I'm new here and have invested a mere fraction of the
> TK> hours in this project that y'all have. So don't hesitate to pat me on
> TK> the head and say "Shhhh! The grown-ups are speaking." :-)
>
> Nah, that would be condescending.  We do passive-aggressive :)
>
> Seriously, your feedback is appreciated.  You have invested a lot of
> time fighting features and jargon that are implied or learned for many
> of the Gnus regulars, so your experience helps us.  If we don't respond
> or fix things quickly (speaking for myself) it's for lack of time, not
> desire to help.
>
> Ted
>

While you're here Ted, I have had to stop using spam-split. Possibly you
can see what is wrong with this?


   (setq spam-use-bogofilter t)
   (setq spam-use-spamassassin t)
   
   (require 'spam)
   (spam-initialize)
   
   (setq nnimap-inbox '("INBOX"))
   (setq nnimap-split-fancy '(| (: spam-split) "INBOX"))
   (setq nnimap-split-methods 'nnimap-split-fancy)
   (setq nnimap-split-download-body t)

I then set the spam settings using G c on the groups I want spam-split
to process.

I get some form of sequencep errors I think it is. Sorry to be vague - I
cant post a backtrace right now but can again tomorrow if you're
interested in taking a look.

I dont *think* there is anything wrong with the above but few people
have offered up other split examples using spam-split to compare it
against. I am guessing my quoting is wrong but its hard to pinpoint from
the manual examples.

regards

r.

Re: Error in (info "(gnus)Splitting Mail")

[Tommy Kelly]

Richard, small and possibly irelevant point. You said:

>    (setq nnimap-inbox '("INBOX"))
> ...
> I get some form of sequencep errors I think it is.

My equivalent is:

(setq nnimap-inbox "INBOX")

I believe that nnimap-innbox can nowadays be a list, but I don't think it
was always so (the docs certainly implied, at one point, that there could
be only one group). Since you only have a single entry there it might be
worth trying it as a simple string. 

Tommy

Re: Error in (info "(gnus)Splitting Mail")

[Richard Riley]

Tommy Kelly <tommy.kelly@verilab.com> writes:

> Richard, small and possibly irelevant point. You said:
>
>>    (setq nnimap-inbox '("INBOX"))
>> ...
>> I get some form of sequencep errors I think it is.
>
> My equivalent is:
>
> (setq nnimap-inbox "INBOX")
>
> I believe that nnimap-innbox can nowadays be a list, but I don't think it
> was always so (the docs certainly implied, at one point, that there could
> be only one group). Since you only have a single entry there it might be
> worth trying it as a simple string. 
>
> Tommy

Yes, I had tried the simpler form too. I'm wondering if anyone is using
spam-split with imap at this stage.  I have the same setup for nnmail -
just a different default split destination and that seems ok.

Re: Error in (info "(gnus)Splitting Mail")

[Richard Riley]

Štěpán Němec <stepnem@gmail.com> writes:

> Ted Zlatanov <tzz@lifelogs.com> writes:
>
>> On Sat, 18 Dec 2010 16:33:47 -0600 Tommy Kelly <tommy.kelly@verilab.com> wrote: 
>>
>> TK> The person who wrote the stuff, understands it in intense
>> TK> detail without the documentation, and therefore has little or no need
>> TK> of the documentation thinks the documentation is more readable as it is.
>>
>> TK> The person who is trying to learn the stuff, and must rely heavily on
>> TK> said documentation thinks the documentation would be more readable with a
>> TK> a minor fix.
>>
>> Štěpán was noting a minor issue: if mail comes from majordomoXcom (any
>> single character) instead of majordomo.com, it will match that splitting
>> rule.  I don't think Štěpán suggested it would make the documentation
>> more readable to fix that.
>
> I think Tommy's and Lars's notion of "readable" is slightly incompatible
> here. In my opinion the error (yes, IMO it _is_ an error) should be
> fixed at least for the reason this whole thread started -- it induced
> major confusion for Tommy, who, upon seeing the unescaped dot, began to
> wonder if there was some magic going on with the fancy splitting rules.
>
> (Yes, it's funny to have discussions about things like this, but not
> having sloppy examples in the docs is precisely the way to prevent such
> discussions.)
>
>   Štěpán

The splitting docs are confusing - issues constantly crop up in #emacs
as well as here and in the nognus mailing list. Possibly more simple
examples would help before they start rolling into what is pretty
advanced elisp where quotes and escaping play an important part.

Re: nnimap-split-fancy

[Eric S Fraga]

Tommy Kelly <tommy.kelly@verilab.com> writes:

[...]

> Yeah, me too. For example: why does the . in domain1.com require the
> "\\" when included in the fancy split approach, but no toothpicks at all
> are needed when the same string is included in the normal (non fancy)
> approach?

Well, because "." matches any character, including "."...

-- 
: Eric S Fraga (GnuPG: 0xC89193D8FFFCF67D) in Emacs 23.2.1 + No Gnus v0.11

Re: nnimap-split-fancy

[Adam Sjøgren]

On Fri, 17 Dec 2010 09:09:13 -0600, Tommy wrote:

> Thanks Adam, but that's my bad. I omitted the closing quotes as I
> created my email message. The original .gnus was correct (font and
> indentation would have shown up that error a mile away).

Always copy/paste and change as little as possible before posting -
sometimes the devil is in the detail.

> Yeah, me too. For example: why does the . in domain1.com require the
> "\\" when included in the fancy split approach, but no toothpicks at all
> are needed when the same string is included in the normal (non fancy)
> approach?

I don't use it enough to care to remember, so I test when I need it.


  Best regards,

    Adam

-- 
 "Accept the mystery!"                                        Adam Sjøgren
                                                         asjo@koldfront.dk

nnimap-split-fancy

[Tommy Kelly]

I may have to propose info "6.3.6 Fancy Mail Splitting" for the winner
of "Most Difficult To Follow Instructions 2010", but let's see. As a
start, does  the following make sense? The intent is to put emails from
domain1.com and domain2.com into, respectively, groups
emails-from-domain1 and emails-from-domain2, and to leave all others in
INBOX. 

(setq gnus-secondary-select-methods
      '(
	(nnimap "work"
		(nnimap-address "imap.gmail.com")
 	 	(nnimap-stream ssl)
		(nnimap-server-port 993)
		(nnimap-inbox "INBOX")
		(nnimap-split-methods 'nnmail-split-fancy)
		(nnmail-split-fancy 
		 ( |
		   ("from" ".*domain1\\.com" "emails-from-domain1)
		   ("from" ".*domain2\\.com" "emails-from-domain2)
		   "INBOX" "")
		  )
		)
      )

If it does make sense, well it doesn't work.

How about an example from someone who groks this stuff. What would the
fancy splitting look like to put into group "jabba-hates-luke", all
messages from jabba@thehutt.com with "Luke is annoying" in the subject? 

And then, other questions about the info:

* "This variable has the format of a split. A split is ..."

Is a "split" a well-defined lisp thing? If not, isn't that definition
nothing more than the statement: "This variable has the format of a
possibly recursive thingy"?  

* "Here are the possible split syntaxes: group ..."

This is very confusing given that "junk" and "nil" are also listed as
options later. It makes it look like group is another symbol that can be
provided as a split. But then:


* "junk  If the split is the symbol junk ..." 
and
* "nil   If the split is nil it is ignored"

What would it look like for the split to *be* the symbol junk?
And an example of a nil split would be helpful.


* "(! func split)  If the split is a list, and the first element is !,
  then split will be processed, and func will be called as a function
  with the result of split as argument..."

Now I'm not showing italics, and so in the original (with-italics) the difference
between *the* split, and its internal split (i.e. third item in the
list) is clearer. But it would be 
a helluva lot easier to read if "split" wasn't being used in two
different senses.

Also, what is "the result of split"? Is it a split? A group name? What?

Note -- I am not au fait with lisp, so maybe the above are all stupid
newb questions, but I'm kinda guessing not. 

thanks,
Tommy