commons-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Eric Pugh" <>
Subject RE: [configuration] problems with commas in string properties
Date Mon, 01 Mar 2004 11:41:05 GMT

Excellently written email!  If you look at some of the posts to
[configuration] over the past two weeks, you'll see us discussing this.  In
a nutshell, different implementations of Configuration dealt with things
differently.  We made some changes to use AbstractConfiguration, which meant
the "," issue came to the fore front for some of the configurations.

I am somewhat loath to change the specification for .properties file prior
to 1.0.  The reason being that the main codebase for Configuration came from
Turbine and Fulcrum, where there are MANY users who have configurations set
up using:


However, I agree, that I think that the right way is

So, what I am going to suggest, for a 1.0 release of Configuration (which is
blocking the 1.0 release for Fulcrum and 2.4 release for Turbine) is that we
keep the "," logic the way it is, but add the ability to change what the
"list delimiter" is.  So you can say:

setListDelimiter(null) // no delimiter
setListDelimiter(";") // change it to ";"

Then when 1.0 is released, we can then start changing the API.  There are a
lot of good ideas for the future of [configuration], but we need to release
what we have...


> -----Original Message-----
> From: Basil James Whitehouse III
> []
> Sent: Sunday, February 29, 2004 3:23 PM
> To:;
> Subject: [configuration] problems with commas in string properties
> I've been using a commons-configuration nightly build from
> February 15,
> 2004 rather successfully.  Recently I have run into a problem with
> strings containing a comma "," in them.
> I'm using XML to store properties, here's an example of the
> XML that is
> causing me grief.
> <configuration>
>     <message-text>A message, with a comma.</message-text>
> <configuration>
> When I use .getString( "message-text" ) the resulting
> property value is
> "A message", whereas I expected the full string.  If the comma is
> escaped (\,) then the resulting value is as expected ("A
> message, with a
> comma.").  While escaping a string isn't that difficult, nor
> uncommon,
> for a developer, it's not something I'd want the lay person
> to have to
> do, and it's easy to forget that it needs to be escaped.  BTW:  this
> particular bit of configuration is going to be done by an
> average user
> who is competent with XML structures, but not much beyond that.
>  From looking at the source I figured out that the comma is used to
> create a List, and I could in fact recreate the text (if the comma is
> not escaped) by running though the list and adding back in
> the commas.
> While I've only glanced at the source, this doesn't seem like
> the best
> approach since if the token changes in the future, or there are
> additional tokens added (like a semi-colon), then I'd have to
> figure out
> what token was used.  It also seems like a List would be
> returned if the
> same property was defined twice in the same file, which I wouldn't be
> able to tell the difference that and a string with a comma in it.  As
> well right now I'm going to assume that there's only going to be one
> message, but I can think of future cases where I might find multiple
> property definitions handy, but also want a string with
> commas in it, in
> the same file.
> So I guess my question is, can this behavior be changed?  If
> .getString
> is used could it return the whole string for the given
> property, and if
> .getList was used could it parse the property then and give
> it back as a
> List?
> Looking at the source I think I've found the bit that does the string
> parsing, AbstractConfiguration.addProperty calls processString which
> parses out a List based on commas.  My cursory glance made me
> think that
> processString could be called in getList, and hence getString would
> return the complete string, but since I don't know how multiple
> properties with the same name (thus creating a list) are represented
> this may not be a correct guess.  I don't presume to know all the
> details of the source, nor driving factors to make these
> decisions, so
> take my code commentary with some salt, but please consider
> this feature
> request.
> Let me know if I can provide any other information, I'd even
> be willing
> to create a test case to support this feature.
> Responses to my email would be appreciated since I'm not
> subscribed to
> the list.
> Thanks,
> Jamie.
> ---------------------------------------------------------------------
> To unsubscribe, e-mail:
> For additional commands, e-mail:

To unsubscribe, e-mail:
For additional commands, e-mail:

View raw message