CSC/ECE 517 Fall 2007/wiki1 3 b6: Difference between revisions

From Expertiza_Wiki
Jump to navigation Jump to search
No edit summary
No edit summary
 
(16 intermediate revisions by the same user not shown)
Line 1: Line 1:
Example use of Currying in Ruby by Ben Parees
== Example use of Currying in Ruby by Ben Parees ==


The following pair of programs demonstrates the value of currying functions in Ruby.  The program itself implements a simple account balance management tool which allows for withdrawals and deposits as specified by the user.  In the first program, a Logger class is defined with a single function, "logMessage" which takes an integer log level parameter, and the message to be logged.  Depending on the log level specified, and the overall logging level set in the Logger class, the message may or may not be displayed to the userIf the message is to be displayed, it will be prefixed with a string indicating the message's relative importance.
This is an original work, the sample code provided was created entirely by the author of this pageReferences are provided to offer additional background on the concept of currying.


The purpose of this class is to allow programmers to easily reconfigure their program to provide varying degrees of debugging output without changing code.  They can simply change the log level that is configured, and automatically filter out messages which are below that log level.  In addition, messages are automatically tagged with a string indicating their severity, which makes analyzing logs much easier.
 
== Definition of Function Currying ==
 
Currying refers to a programming technique in which the number of parameters to a function are reduced by creating a new function based on the original, but fixing some of the original arguments.  That is, if we have a function ''f(x,y,z)'', we can define a new function ''g(x)'' which is defined as ''f(x,1,2)''.
 
Now ''f(x,1,2)==g(x)'', allowing us to invoke the simpler g(x) rather than ''f(x,1,2)''.  To curry a function means to create the simplified function with fixed arguments, from a function with more arguments.
 
 
== Benefits of Performing Function Currying ==
 
Currying has several benefits.
<ol>
<li>It simplifies writing code by reducing the number of arguments that must be typed each time</li>
<li>It reduces the risk of error because there are fewer arguments for which values could be incorrect</li>
<li>It can enhance code readability by defining multiple functions with descriptive names rather than a single function with obscure arguments</li>
</ol>
 
 
== Currying in Ruby ==
 
Ruby provides for function currying using the lambda operator.  The lambda operator allows the programmer to define a new function.  To curry'' g(x)'' from ''f(x,1,2)'' in Ruby would be written as: ''g=lambda {|x| f(x,1,2)}''
This would assign variable g to be a new function which invokes f() with the final two arguments fixed as 1 and 2 respectively.
 
 
=== Example Program in Ruby ===
 
The following pair of programs demonstrates the value of currying functions in Ruby.  The program itself implements a simple account balance management tool which allows for withdrawals and deposits as specified by the user. 
 
In the first program, a Logger class is defined with a single function, "logMessage" which takes an integer log level parameter, and the message to be logged.  Depending on the log level specified, and the overall logging level set in the Logger class, the message may or may not be displayed to the user.  If the message is to be displayed, it will be prefixed with a string indicating the message's relative importance. 
 
For example, logMessage(0,"starting withdrawal") will print a debugging message as:
<pre>DEBUG: starting withdrawal</pre>
 
In addition, this message will only be displayed if the logLevel is set to 0 or less.  The more important the message, the higher the log level that should be used, to ensure it is displayed to the user.  Since debug messages are often unimportant, setting the logLevel to 1 would prevent all debug messages from being shown to the user while still showing the user normal output and warning/error messages.  When debugging the program, the programmer can set the logLevel to 0 to enable the output of these debug messages.
 
The purpose of this Logger class is to allow programmers to easily reconfigure their program to provide varying degrees of debugging output without changing code.  They can simply change the log level that is configured, and automatically filter out messages which are below that log level.  In addition, messages are automatically tagged with a string indicating their severity, which makes analyzing logs much easier. Unfortunately, the logMessage method is fairly cumbersome due to the requirement that the programmer specify the correct log level at each invocation.
 
 
==== Sample Program without Currying ====


Here is the initial version of the program:
Here is the initial version of the program:


<pre>
class Logger
class Logger
         @logLevel
         @logLevel
Line 13: Line 52:
                 @logLevel=n
                 @logLevel=n
         end
         end
         def logMessage(n,msg)
         def logMessage(n,msg)
                 if(n>=@logLevel)
                 if(n>=@logLevel)
Line 20: Line 58:
                                 when 1 then ''
                                 when 1 then ''
                                 when 2 then 'WARNING: '
                                 when 2 then 'WARNING: '
                                 when 3 then 'ERROR: '
                                 when 3 then 'ERROR: '                              
                                else "hmmmmm"
                         end
                         end
                         puts prepend+msg
                         puts prepend+msg
Line 37: Line 74:
                 logger.logMessage(0,"initialized account with balance of #{n}");
                 logger.logMessage(0,"initialized account with balance of #{n}");
         end
         end
         def getAmount
         def getAmount
           @logger.logMessage(0,"enter getAmount method")
           @logger.logMessage(0,"enter getAmount method")
Line 69: Line 104:
           @logger.logMessage(0,"new balance is #{balance}")
           @logger.logMessage(0,"new balance is #{balance}")
         end
         end
end
end
action="d"
action="d"
logger=Logger.new(0)
logger=Logger.new(0)
account=Account.new(100,logger)
account=Account.new(100,logger)
while action!="q"
while action!="q"


Line 90: Line 122:
         end
         end
end
end
</pre>




==== Limitations of the Initial Program ====


In the initial form of the program, the programmer must specify the log level parameter for each trace statement they wish to create.  This approach has two drawbacks:
In the initial form of the program, the programmer must specify the log level parameter for each trace statement they wish to create.  This approach has two drawbacks:
1) The programmer must know what log level corresponds to what level of trace (0=debug, 1=info, 2=warning, 3=error, etc).  If they make a mistake in their numbering, the trace will not be correct.
<br>
2) The code is hard to read since it is not obvious just by looking at a piece of code, whether a message is going to be printed as a debug statement, informational statement, warning, etc.
<ol>
<li>The programmer must know what log level corresponds to what level of trace (0=debug, 1=info, 2=warning, 3=error, etc).  If they make a mistake in their numbering, the trace will not be correct.</li>
<li>The code is hard to read since it is not obvious just by looking at a piece of code, whether a message is going to be printed as a debug statement, informational statement, warning, etc.</li>
</ol>
 


==== Enhancing the program via Currying ====


Now the same program, but with 4 new methods defined by currying the original logMessage method.  The new methods are:
Now the same program, but with 4 new methods defined by currying the original logMessage method.  The new methods are:
1) debug - print a debug level message
<ol>
2) print - print an informational message
<li> debug - print a debug level message by currying with logMessage(0,msg) </li>
3) warn - print a warning level message
<li> print - print an informational message by currying with logMessage(1,msg) </li>
4) error - print an error level message
<li> warn - print a warning level message by currying with logMessage(2,msg)</li>
<li> error - print an error level message by currying with logMessage(3,msg)</li>
</ol>
The actual currying is done in the Logger class, and the invocation of these methods can be seen in the body of the main program and the Account class.
 
Here is the section of code which performs the currying:
<pre>
def initialize(n)
                @logLevel=n
                @debug=lambda{|msg| logMessage(0,msg)}
                @print=lambda{|msg| logMessage(1,msg)}
                @warn= lambda{|msg| logMessage(2,msg)}
                @error=lambda{|msg| logMessage(3,msg)}
end
</pre>
 
 
==== Improved Version of the Program ====


The actual currying is done in the Logger class, and the invocation of these methods can be seen in the body of the main program and the Account class.


And here is the entire program rewritten to utilize the new methods.
<pre>
class Logger
class Logger
         @logLevel
         @logLevel
Line 128: Line 186:
                                 when 2 then 'WARNING: '
                                 when 2 then 'WARNING: '
                                 when 3 then 'ERROR: '
                                 when 3 then 'ERROR: '
                                else "hmmmmm"
                         end
                         end
                         puts prepend+msg
                         puts prepend+msg
Line 194: Line 251:


end
end


action="d"
action="d"
logger=Logger.new(0)
logger=Logger.new(0)
account=Account.new(100,logger)
account=Account.new(100,logger)


while action!="q"
while action!="q"
Line 215: Line 270:
         end
         end
end
end
</pre>
== Conclusion ==
Notice how much easier it is to follow the flow of the sample program now through the clear labeling of debug/print/warning/error statements.  It is very clear which output will be shown to the user and which is only provided for debugging purposes. 
In addition, observe how simple it would be to add a new trace statement at any log level simply by invoking the desired method rather than having to look up the appropriate log level value.
In general, currying allows programmers to write stricter code which is less likely to contain misspecified arguments, and be easier to read by virtue of not requiring the reader to know what each argument value means.  It is a very valuable technique for enhancing code reliability and readability.
== References ==
#[http://www.amazon.com/Programming-Ruby-Pragmatic-Programmers-Second/dp/0974514055 Programming Ruby: The programmatic programmer’s guide]
#[http://www.cs.nott.ac.uk/~gmh/faq.html#currying Discussion of the origins and definition of currying]
#[http://moonbase.rydia.net/mental/blog/programming/currying-in-ruby A blog entry about the concept of currying]
#[http://en.wikipedia.org/wiki/Currying Wikipedia entry on currying]
#Lecture notes provided by Dr. Gehringer about Ruby and Currying
== See Also ==
#[http://rightfootin.blogspot.com/2006/09/spicey-curry-in-your-ruby.html Code to make currying easier in Ruby]
--[[User:Bmparees|Bmparees]] 00:48, 12 September 2007 (EDT)

Latest revision as of 04:16, 16 September 2007

Example use of Currying in Ruby by Ben Parees

This is an original work, the sample code provided was created entirely by the author of this page. References are provided to offer additional background on the concept of currying.


Definition of Function Currying

Currying refers to a programming technique in which the number of parameters to a function are reduced by creating a new function based on the original, but fixing some of the original arguments. That is, if we have a function f(x,y,z), we can define a new function g(x) which is defined as f(x,1,2).

Now f(x,1,2)==g(x), allowing us to invoke the simpler g(x) rather than f(x,1,2). To curry a function means to create the simplified function with fixed arguments, from a function with more arguments.


Benefits of Performing Function Currying

Currying has several benefits.

  1. It simplifies writing code by reducing the number of arguments that must be typed each time
  2. It reduces the risk of error because there are fewer arguments for which values could be incorrect
  3. It can enhance code readability by defining multiple functions with descriptive names rather than a single function with obscure arguments


Currying in Ruby

Ruby provides for function currying using the lambda operator. The lambda operator allows the programmer to define a new function. To curry g(x) from f(x,1,2) in Ruby would be written as: g=lambda {|x| f(x,1,2)} This would assign variable g to be a new function which invokes f() with the final two arguments fixed as 1 and 2 respectively.


Example Program in Ruby

The following pair of programs demonstrates the value of currying functions in Ruby. The program itself implements a simple account balance management tool which allows for withdrawals and deposits as specified by the user.

In the first program, a Logger class is defined with a single function, "logMessage" which takes an integer log level parameter, and the message to be logged. Depending on the log level specified, and the overall logging level set in the Logger class, the message may or may not be displayed to the user. If the message is to be displayed, it will be prefixed with a string indicating the message's relative importance.

For example, logMessage(0,"starting withdrawal") will print a debugging message as:

DEBUG: starting withdrawal

In addition, this message will only be displayed if the logLevel is set to 0 or less. The more important the message, the higher the log level that should be used, to ensure it is displayed to the user. Since debug messages are often unimportant, setting the logLevel to 1 would prevent all debug messages from being shown to the user while still showing the user normal output and warning/error messages. When debugging the program, the programmer can set the logLevel to 0 to enable the output of these debug messages.

The purpose of this Logger class is to allow programmers to easily reconfigure their program to provide varying degrees of debugging output without changing code. They can simply change the log level that is configured, and automatically filter out messages which are below that log level. In addition, messages are automatically tagged with a string indicating their severity, which makes analyzing logs much easier. Unfortunately, the logMessage method is fairly cumbersome due to the requirement that the programmer specify the correct log level at each invocation.


Sample Program without Currying

Here is the initial version of the program:

class Logger
        @logLevel
        attr_accessor :logLevel
        def initialize(n)
                @logLevel=n
        end
        def logMessage(n,msg)
                if(n>=@logLevel)
                        prepend=case n
                                when 0 then 'DEBUG: '
                                when 1 then ''
                                when 2 then 'WARNING: '
                                when 3 then 'ERROR: '                                
                        end
                        puts prepend+msg
                end

        end
end
class Account
        @balance
        @logger
        attr_reader :balance
        def initialize(n,logger)
                @balance=n
                @logger=logger
                logger.logMessage(0,"initialized account with balance of #{n}");
        end
        def getAmount
          @logger.logMessage(0,"enter getAmount method")
          amount=gets.chomp!.to_f
          @logger.logMessage(0,"Got amount #{amount}")
          if(amount<0)
                @logger.logMessage(3,"Amount is less than zero, amount will be set to zero")
                amount=0
          end
          return amount
        end

        def withdraw
          @logger.logMessage(0,"enter withdraw method")
          @logger.logMessage(1,"How much would you like to withdraw?")
          amount=getAmount
          @logger.logMessage(0,"withdrawing amount of #{amount}")
          @balance=@balance-amount
          if(@balance<0)
                @logger.logMessage(2,"Account is overdrawn")
          end
          @logger.logMessage(0,"new balance is #{@balance}")
        end

        def deposit
          @logger.logMessage(0,"enter depost method")
          amount=getAmount
          @logger.logMessage(0,"depositing amount of #{amount}")
          @balance=@balance+amount
          @logger.logMessage(0,"new balance is #{balance}")
        end
end
action="d"
logger=Logger.new(0)
account=Account.new(100,logger)
while action!="q"

        logger.logMessage(1,"Balance=#{account.balance}, would you like to withdraw, deposit, or quit [w/d/q]?")

        action=gets.chomp!
        logger.logMessage(0,"user chose action #{action}")
        if action=="w"
          account.withdraw
        elsif action=="d"
          account.deposit
        elsif action=="q"
          logger.logMessage(1,"goodbye")
        end
end


Limitations of the Initial Program

In the initial form of the program, the programmer must specify the log level parameter for each trace statement they wish to create. This approach has two drawbacks:

  1. The programmer must know what log level corresponds to what level of trace (0=debug, 1=info, 2=warning, 3=error, etc). If they make a mistake in their numbering, the trace will not be correct.
  2. The code is hard to read since it is not obvious just by looking at a piece of code, whether a message is going to be printed as a debug statement, informational statement, warning, etc.


Enhancing the program via Currying

Now the same program, but with 4 new methods defined by currying the original logMessage method. The new methods are:

  1. debug - print a debug level message by currying with logMessage(0,msg)
  2. print - print an informational message by currying with logMessage(1,msg)
  3. warn - print a warning level message by currying with logMessage(2,msg)
  4. error - print an error level message by currying with logMessage(3,msg)

The actual currying is done in the Logger class, and the invocation of these methods can be seen in the body of the main program and the Account class.

Here is the section of code which performs the currying:

def initialize(n)
                @logLevel=n
                @debug=lambda{|msg| logMessage(0,msg)}
                @print=lambda{|msg| logMessage(1,msg)}
                @warn= lambda{|msg| logMessage(2,msg)}
                @error=lambda{|msg| logMessage(3,msg)}
end


Improved Version of the Program

And here is the entire program rewritten to utilize the new methods.

class Logger
        @logLevel
        @debug
        @print
        @warn
        @error
        attr_accessor :logLevel
        def initialize(n)
                @logLevel=n
                @debug=lambda{|msg| logMessage(0,msg)}
                @print=lambda{|msg| logMessage(1,msg)}
                @warn= lambda{|msg| logMessage(2,msg)}
                @error=lambda{|msg| logMessage(3,msg)}
        end

        def logMessage(n,msg)
                if(n>=@logLevel)
                        prepend=case n
                                when 0 then 'DEBUG: '
                                when 1 then ''
                                when 2 then 'WARNING: '
                                when 3 then 'ERROR: '
                        end
                        puts prepend+msg
                end

        end

        def debug(msg)
                @debug.call(msg)
        end
        def print(msg)
                @print.call(msg)
        end
        def warn(msg)
                @warn.call(msg)
        end
        def error(msg)
                @error.call(msg)
        end


end

class Account
        @balance
        @logger
        attr_reader :balance
        def initialize(n,logger)
                @balance=n
                @logger=logger
                logger.debug("initialized account with balance of #{n}");
        end


        def getAmount
          @logger.debug("enter getAmount method")
          amount=gets.chomp!.to_f
          @logger.debug("Got amount #{amount}")
          if(amount<0)
                @logger.error("Amount is less than zero, amount will be set to zero")
                amount=0
          end
          return amount
        end

        def withdraw
          @logger.debug("enter withdraw method")
          @logger.print("How much would you like to withdraw?")
          amount=getAmount
          @logger.debug("withdrawing amount of #{amount}")
          @balance=@balance-amount
          if(@balance<0)
                @logger.warn("Account is overdrawn")
          end
          @logger.debug("new balance is #{@balance}")
        end

        def deposit
          @logger.debug("enter depost method")
          amount=getAmount
          @logger.debug("depositing amount of #{amount}")
          @balance=@balance+amount
          @logger.debug("new balance is #{balance}")
        end

end

action="d"
logger=Logger.new(0)
account=Account.new(100,logger)

while action!="q"

        logger.print("Balance=#{account.balance}, would you like to withdraw, deposit, or quit [w/d/q]?")

        action=gets.chomp!
        logger.debug("user chose action #{action}")
        if action=="w"
          account.withdraw
        elsif action=="d"
          account.deposit
        elsif action=="q"
          logger.print("goodbye")
        end
end


Conclusion

Notice how much easier it is to follow the flow of the sample program now through the clear labeling of debug/print/warning/error statements. It is very clear which output will be shown to the user and which is only provided for debugging purposes.

In addition, observe how simple it would be to add a new trace statement at any log level simply by invoking the desired method rather than having to look up the appropriate log level value.

In general, currying allows programmers to write stricter code which is less likely to contain misspecified arguments, and be easier to read by virtue of not requiring the reader to know what each argument value means. It is a very valuable technique for enhancing code reliability and readability.

References

  1. Programming Ruby: The programmatic programmer’s guide
  2. Discussion of the origins and definition of currying
  3. A blog entry about the concept of currying
  4. Wikipedia entry on currying
  5. Lecture notes provided by Dr. Gehringer about Ruby and Currying

See Also

  1. Code to make currying easier in Ruby

--Bmparees 00:48, 12 September 2007 (EDT)