Feature #12913: A way to configure the default maximum width of pp

• Status: Rejected
• Priority: Normal
• Assignee: akr (Akira Tanaka)
• Target version: -

[mame (Yusuke Endoh)]

How about having an easy way to configure the maximum width of a line of pp output?
Currently, pp accepts the maximum width as an optional argument:

pp(big_array, $>, 120)

However, this is obviously too long for a useful debugging-purpose method like pp. Even worse, we must add the fragment ", $>, 120" to all calls to pp. I don't feel this is reasonable.

The patch attached provides PP.default_maxwidth= and PP.default_maxwidth, which can be used to configure the default setting of the maxwidth.

PP.default_maxwidth = 1
pp([1, 2, 3])
#=> [1,
#    2,
#    3]

Akr-san, what do you think?

[akr (Akira Tanaka)]

I think the columns obtained by IO#winsize of io/console is better default now.

I'm not sure you are satisfied with it, though.

[mame (Yusuke Endoh)]

Akr-san, thank you for your reply.

I agree that it is a better default than the fixed 79.

However, I'd still like to have PP.default_maxwidth= too.
My motivation is to show abstract syntax tree (in my article).
In such a case, I think that "break as far as possible" would be useful.

Thank you for your consideration,

[akr (Akira Tanaka)]

• Status changed from Open to Feedback

Yusuke Endoh wrote:

In such a case, I think that "break as far as possible" would be useful.

"break as far as possible" is opposite to pp.rb's intent:
"don't break as far as possible (in screen width limitation)".

I think a global configuration to choose different intent for single API is not a good idea.

pp (including pretty_inspect and PP.pp) method is used in many gems.
I think the intent of most of them is "don't break as far as possible".
PP.default_maxwidth = 1 breaks the intent.

I recommend that you define a method to call PP.pp with width = 1.

[mame (Yusuke Endoh)]

Akira Tanaka wrote:

pp (including pretty_inspect and PP.pp) method is used in many gems.

Note that my proposal is only for PP; the scope does not include pretty_inspect.

PP is a debugging-purpose API, so it is rarely used in a released code. So, we don't have to consider the case where PP is used in multiple modules simultaneously, I think. Also, because of debugging-purpose, it would be good to prioritize usability than safety.

[akr (Akira Tanaka)]

I think pp is not only for debug.
I guess someone use it for logging, for example.

So, I think global configuration is dangerous,

[mame (Yusuke Endoh)]

• Status changed from Feedback to Rejected

Okay, I respect your decision. Thank you for your time.

[nobu (Nobuyoshi Nakada)]

What about this?

• affects PP.pp and Kernel#pp only
• try console window size, COLUMNS environment variable, then old good 80

diff --git a/lib/pp.rb b/lib/pp.rb
indiff --git a/lib/pp.rb b/lib/pp.rb
index 85401c8aa6..a364d0fc2a 100644
--- a/lib/pp.rb
+++ b/lib/pp.rb
@@ -68,7 +68,7 @@
   # If +width+ is omitted, 79 is assumed.
   #
   # PP.pp returns +out+.
-  def PP.pp(obj, out=$>, width=79)
+  def PP.pp(obj, out=$>, width=PP.width_for(out))
     q = PP.new(out, width)
     q.guard_inspect_key {q.pp obj}
     q.flush
@@ -91,6 +91,15 @@
   def PP.mcall(obj, mod, meth, *args, &block)
     mod.instance_method(meth).bind(obj).call(*args, &block)
   end
+
+  def PP.width_for(out)
+    begin
+      require 'io/console'
+      _, width = out.winsize
+    rescue LoadError, NoMethodError, Errno::ENOTTY
+    end
+    (width || ENV['COLUMNS']&.to_i&.nonzero? || 80) - 1
+  end
   # :startdoc:

   @sharing_detection = false
@@ -549,9 +558,9 @@
   # prints arguments in pretty form.
   #
   # pp returns argument(s).
-  def pp(*objs)
+  def pp(*objs, out: $>, width: PP.width_for(out))
     objs.each {|obj|
-      PP.pp(obj)
+      PP.pp(obj, out, width)
     }
     objs.size <= 1 ? objs.first : objs
   end

[akr (Akira Tanaka)]

nobu (Nobuyoshi Nakada) wrote:

What about this?

• affects PP.pp and Kernel#pp only
• try console window size, COLUMNS environment variable, then old good 80

It seems fine.