Re: [Yaml-core] Evans and Ingerson to meet in 13 hours

[Clark C . E. <cc...@cl...>]

On Mon, May 20, 2002 at 02:09:59AM -0400, Oren Ben-Kiki wrote:
| I slept over this... and I think *true is a bad idea.
| I see the advantage: being able to write booleans in a 
| readable way. But I think this is the wrong way to achieve it.

The rationale for this emerged using how I currently
implement enumerations within my data.  For example:

--- !http://clarkevans.com/^task-list
 - 
   name: Read-email
   status: &001 !^status Continuous
   hours: 1
 - 
   name: Write Congress person
   status: &002 !^status In-Progress
   hours: 2
 - 
   name: Read news
   status: *001 # Continuous
   hours: 3

Thus, the anchor here is a typed node and later
occurances of the enumeration is done via alias
with a comment.  This is a very clean way to emit.
The actual enumeration can be put in a schema for
validation and/or visual editors.

| If we ever want to use the same trick for other
| enum-like types, this could be a problem.

Well, *bool is a natural extension of the above
mechanism if you ask me...

| ---
| - *true
| - &true bloop
| - *true

This one is rather clear, the user has re-defined
what *true means.  Not a problem.  In perl, "int"
is a built-in; however, you can shadow it.

x = int("3")   # converts 3 to an integer type
int = "I am a string now."
y = int("3")   # syntax error, int is a string.

This may seem problematic, but it really isn't.  Keywords
can be added without breaking old code... 

| - Additional complexity.
| 
| The above creates a new mechanism, "predefined anchors". This means that it
| is no longer sufficient to specify the type families to be able to fully
| grok a YAML file. We would also have to specify a set of pre-defined anchors
| the document relies on.

Well, those pre-defined can be in the YAML spec.  If we
increase the number of pre-defined anchors, we will have
to increment the YAML version (the minor version).

| - Slippery slope.
| 
| Where does it stop? *pi, *e and *i? *noon, *midnight and *tea-time? *gpl and
| *bsd? Please. Let's not start down this road.

Ok.

| canonical, explicit: 0|1
| implicit: :[tT]rue|:TRUE|:[Ff]alse|:FALSE
| implicit: :[Yy]es|:YES|:[Nn]o|:NO
| 
| examples:
| 
| trues:
|  - !bool 1
|  - :true
|  - :True
|  - :TRUE
| falses:
|  - !bool 0
|  - :false
|  - :False
|  - :FALSE

Ok.  This isn't bad, although I don't particulary like
the use of the colon (:) as we already have given this
meaning.

Best,

Clark

-- 
Clark C. Evans                   Axista, Inc.
http://www.axista.com            800.926.5525
XCOLLA Collaborative Project Management Software

"There lies before us, if we choose, continued
progress in happiness, knowledge, wisdom. Shall
we, instead, choose death, because we cannot forget
our quarrels?" -- Albert Einstein and Bertrand
Russell in 1955 paper regarding nuclear weapons.
http://www.noradiation.org/caldicott/book-02.html