refdb-users

[Refdb-users] php web addnote functional

[Matt P. <mat...@ut...>]

Hey folks,

I have finally (after demonstrating to myself many times over that I'm
 a terrible programmer) gotten a barely adequate version of a web 'add
note' page working.  You can find it here:
http://www.derailleur.org/webref/index.php?format=risx&field=CK&searchText=Garthwaite2002

a couple of serious problems:

-searching by author is entirely broken.  THis seems to be an issue
with my refdb installation, as it doesn't work from refbc either.

- searches don't work on regex's -- that is, they only work when the
precise, full field value is entered.  Again this seems to be a
problem with my installation of refdb:  
	refdbc: getref :CK:=Martin2001
	gets a reference, whereas
	refdbc: getref :CK:=Martin 
	does not.

- the 'comment.php' page has no prodection against double-loading, so
it's easy to screw up and addyour comment twice just by hitting the reload button.

- this last leads me to another issue with my refdb installation:  I
can't delete notes.  
e.g.: 
 refdbc: deletenote :NID:>0
 0 datasets deleted, 0 skipped
... and also I periodically lose the capacity to add notes, and have
to rebuild my db to get it back

So there appears to be something seriously wrong with my refdb
install, which I will look into.  But for the meantime, the addnotes
is working ok, please give ocmmetns.!

matt

Re: [Refdb-users] php web addnote functional

[Bruce D'A. <bd...@fa...>]

On Feb 13, 2004, at 7:37 AM, Matt Price wrote:

> I have finally (after demonstrating to myself many times over that I'm
>  a terrible programmer) gotten a barely adequate version of a web 'add
> note' page working

Cool.

I still think Moveable Type's pop-up comments UI is a better model 
though.  Even with one record, I find the notes UI now a little 
confusing (if I have the notes displayed, how do I get rid of them?  
What does "add note" do?).  If there are many records, it seems like 
it'll be more confusing.

Bruce

[Refdb-users] php web addnote functional

[Markus H. <mar...@mh...>]

Hi Matt,

Matt Price writes:
 > Hey folks,
 > 
 > I have finally (after demonstrating to myself many times over that I'm
 >  a terrible programmer) gotten a barely adequate version of a web 'add
 > note' page working.  You can find it here:
 > http://www.derailleur.org/webref/index.php?format=risx&field=CK&searchText=Garthwaite2002
 > 

First of all, thanks for this effort. Much appreciated.

 > a couple of serious problems:
 > 
 > -searching by author is entirely broken.  THis seems to be an issue
 > with my refdb installation, as it doesn't work from refbc either.
 > 

This is related to the next "problem".

 > - searches don't work on regex's -- that is, they only work when the
 > precise, full field value is entered.  Again this seems to be a
 > problem with my installation of refdb:  
 > 	refdbc: getref :CK:=Martin2001
 > 	gets a reference, whereas
 > 	refdbc: getref :CK:=Martin 
 > 	does not.
 > 

As you're running the CVS version, you're ahead of the latest NEWS
file (not yet checked in) which will explain a change in the query
syntax. Queries using '=' or '!=' now default to a literal match, this
is why 
   refdbc: getref :CK:=Martin2001
works for you. If you want a partial or a regexp match, you have to
use the new tilde operators '~' and '!~' instead:
   refdbc: getref :CK:~Martin

So if I understand your problem correctly, this is an "early adopter"
issue. Sorry for the inconvenience.

 > - this last leads me to another issue with my refdb installation:  I
 > can't delete notes.  
 > e.g.: 
 >  refdbc: deletenote :NID:>0
 >  0 datasets deleted, 0 skipped

The deletenote command expects a whitespace-separated list of ID
values, like this:
    refdbc: deletenote 1 3 4-6
The < and > operators are not (yet) implemented.

 > ... and also I periodically lose the capacity to add notes, and have
 > to rebuild my db to get it back
 > 

This is a serious issue that I can't reproduce here. In order to fix
this I seriously need more debug information. If you find the time,
please

- start refdbd with a log level of 7
- start with a fresh database
- use a defined set of notes and add them until addnote fails
- send me the log and the set of notes that caused the failure.

I can't fix things which I can't reproduce. I need these data.

Thanks
Markus

-- 
Markus Hoenicka
mar...@ca...
(Spam-protected email: replace the quadrupeds with "mhoenicka")
http://www.mhoenicka.de

Re: [Refdb-users] php web addnote functional

[Matt P. <mat...@ut...>]

On Fri, Feb 13, 2004 at 10:32:20PM +0100, Markus Hoenicka wrote:
> Hi Matt,
> 
> Matt Price writes:
>  > Hey folks,
>  > 
>  > I have finally (after demonstrating to myself many times over that I'm
>  >  a terrible programmer) gotten a barely adequate version of a web 'add
>  > note' page working.  You can find it here:
>  > http://www.derailleur.org/webref/index.php?format=risx&field=CK&searchText=Garthwaite2002
>  > 
> 
> First of all, thanks for this effort. Much appreciated.
> 
>  > a couple of serious problems:
>  > 
>  > -searching by author is entirely broken.  THis seems to be an issue
>  > with my refdb installation, as it doesn't work from refbc either.
>  > 
> 
> This is related to the next "problem".
> 
>  > - searches don't work on regex's -- that is, they only work when the
>  > precise, full field value is entered.  Again this seems to be a
>  > problem with my installation of refdb:  
>  > 	refdbc: getref :CK:=Martin2001
>  > 	gets a reference, whereas
>  > 	refdbc: getref :CK:=Martin 
>  > 	does not.
>  > 
> 
> As you're running the CVS version, you're ahead of the latest NEWS
> file (not yet checked in) which will explain a change in the query
> syntax. Queries using '=' or '!=' now default to a literal match, this
> is why 
>    refdbc: getref :CK:=Martin2001
> works for you. If you want a partial or a regexp match, you have to
> use the new tilde operators '~' and '!~' instead:
>    refdbc: getref :CK:~Martin
> 
> So if I understand your problem correctly, this is an "early adopter"
> issue. Sorry for the inconvenience.
> 
>  > - this last leads me to another issue with my refdb installation:  I
>  > can't delete notes.  
>  > e.g.: 
>  >  refdbc: deletenote :NID:>0
>  >  0 datasets deleted, 0 skipped
> 
> The deletenote command expects a whitespace-separated list of ID
> values, like this:
>     refdbc: deletenote 1 3 4-6
> The < and > operators are not (yet) implemented.
> 
>  > ... and also I periodically lose the capacity to add notes, and have
>  > to rebuild my db to get it back
>  > 
> 
> This is a serious issue that I can't reproduce here. In order to fix
> this I seriously need more debug information. If you find the time,
> please
> 
> - start refdbd with a log level of 7
> - start with a fresh database
> - use a defined set of notes and add them until addnote fails
> - send me the log and the set of notes that caused the failure.
> 
> I can't fix things which I can't reproduce. I need these data.


Hey Markus, sorry to take so long getting you data, but I'm being
somewhat unsystematic at the moment -- I will do my best to reproduce
the error faithfully and get back to you, but it will likely be a
couple of days at best.  Looking at my mail it sounds rather
plaintive, and I don't mean that actually -- I suspect the problem may
lie, for instance, in a faulty installation ofm ysql sor some such.   

anyway it's my birthday and we have a babysitter so I'm off to the
movies -- more sometime later!

matt
.  
> 
> Thanks
> Markus
> 

Re: [Refdb-users] php web addnote functional

[Markus H. <mar...@mh...>]

Matt Price writes:
 > Hey Markus, sorry to take so long getting you data, but I'm being
 > somewhat unsystematic at the moment -- I will do my best to reproduce
 > the error faithfully and get back to you, but it will likely be a
 > couple of days at best.  Looking at my mail it sounds rather
 > plaintive, and I don't mean that actually -- I suspect the problem may
 > lie, for instance, in a faulty installation ofm ysql sor some such.   
 > 

No, don't take me wrong. I'm just trying to get 0.9.4 out the door,
and I can't do that as long as there are unresolved issues. That's why
I'm pressing a bit, sorry about that.

 > anyway it's my birthday and we have a babysitter so I'm off to the
 > movies -- more sometime later!
 > 

congrats & enjoy!

Markus

-- 
Markus Hoenicka
mar...@ca...
(Spam-protected email: replace the quadrupeds with "mhoenicka")
http://www.mhoenicka.de

[Refdb-users] refdb & '&'

[Markus H. <mar...@mh...>]

Matt Price writes:
 > Hey Markus:
 > 
 > a couple notes:
 > 
 > I may have overstated my problems with refdb earlier.  I think my
 > 'addnote' issues don't actually require rebuilding the database, but
 > rather, just restarting refdbd.  obviously that's way less drastic.
 > 

I could track down the problem with the help of your logs. I can't
explain why in your case restarting refdbd got you around the problem
(it can't, if I understand my code correctly :-). In any case, no one
wouldn't want to restart an app anyway. I think I've fixed the
problem. In brief, adding a dataset or a note involves creating a
temporary citation key if the dataset doesn't specify one. If there is
a problem with a dataset and it is not removed properly from the
database again, this citation key may block the next addnote/addref
command as it has to be unique. That is, this problem occurred only
with a database engine that does not use transactions, only with
datasets that cause a parse error (see below), and only with datasets
that don't specify a citation key. I've fixed the code that removes
partial datasets if you can't do a rollback, and I've also fixed a bug
that swallowed the error message of the XML parser. It now returns for
your dataset:

refdbc: addnote ../pseudotest/data/xnoteset.xml
server error: Parse error at line 50:
not well-formed (invalid token)

Now it should be pretty clear what's going on.

 > As for the note, the stumbling block seems to be the presence of the
 > character '&'.  Do you know why this would be?  being quite ignorant
 > of almost all the technologies involved here, I'm not sure what the
 > issue is.  
 > 

Yep. The ampersand is one of the reserved characters in XML (others
would be < and >). In order to use a literal ampersand in an XML
document, use the entity "&amp;" instead. Same with "&lt;" and "&gt;"
for < and >, respectively.

Thanks again for helping me spot this bug.

regards,
Markus

-- 
Markus Hoenicka
mar...@ca...
(Spam-protected email: replace the quadrupeds with "mhoenicka")
http://www.mhoenicka.de