explain this function to me, lambda confusion

G

globalrev

i have a rough understanding of lambda but so far only have found use
for it once(in tkinter when passing lambda as an argument i could
circumvent some tricky stuff).
what is the point of the following function?

def addn(n):
return lambda x,inc=n: x+inc

if i do addn(5) it returns

return lambda x,inc=n: x+inc

<function <lambda> at 0x01D81830>


ok? so what do i write to make it actually do something. and is the
inc=n necessary i cant do x+n?
 
G

globalrev

and what si the diffrence here:

g = lambda x=5:x*x
g = lambda x:x*x

the first was a mistake to write but it worked
and the x=5 seems to be completely ignored. why? it has no effect at
all?
 
G

globalrev

and what si the diffrence here:

g = lambda x=5:x*x
g = lambda x:x*x

the first was a mistake to write but it worked
and the x=5 seems to be completely ignored. why? it has no effect at
all?

ah wait now i see it has a default kind of then. g() returns 25 while
g(8) returns 64.
 
B

bruno.desthuilliers

i have a rough understanding of lambda but so far only have found use
for it once(in tkinter when passing lambda as an argument i could
circumvent some tricky stuff).
what is the point of the following function?

def addn(n):
return lambda x,inc=n: x+inc

It returns a function that accept one argument and return the result
of the addition of this argument with the argument passed to addn.

FWIW, Python's lambda is just a shortcut to create a very simple
function, and the above code is canonically written as:

def makeadder(n):
def adder(x):
return n + x
return adder
if i do addn(5) it returns
(snip)

<function <lambda> at 0x01D81830>

ok? so what do i write to make it actually do something.

add5 = addn(5)
add5(1)
=> 6
add5(2)
=> 7

add42 = addn(42)
add42(1)
=> 43
and is the
inc=n necessary i cant do x+n?

In this case, it's not. This version does exactly the same thing
AFAICT:

def addn(n):
return lambda x: x+n
 
G

Gabriel Genellina

i have a rough understanding of lambda but so far only have found use
for it once(in tkinter when passing lambda as an argument i could
circumvent some tricky stuff).
what is the point of the following function?

def addn(n):
return lambda x,inc=n: x+inc

lambda is just a shortcut for defining a function without a name.
The above code is the same as:

def addn(n):
def inner(x, inc=n):
return x+inc
return inner

It should be clear now that addn returns a function. addn is a "function
factory": builds functions by request. You ask it "give me a function that
adds 5" and addn returns that function.
if i do addn(5) it returns

<function <lambda> at 0x01D81830>

If you try the other version, you would get:

<function inner at 0x00A3B970>

It's the same thing, except that lambda has no name.
ok? so what do i write to make it actually do something.

adder5 = addn(5)
adder5(3)
-> 8
and is the
inc=n necessary i cant do x+n?

Yes, you can, but there is a subtle difference that's hard to explain, and
in this case it's absolutely irrelevant. Using inc=n "does the right
thing" as it's a bit more efficient too.
 
A

andrej.panjkov

i have a rough understanding of lambda but so far only have found use
for it once(in tkinter when passing lambda as an argument i could
circumvent some tricky stuff).
what is the point of the following function?

def addn(n):
return lambda x,inc=n: x+inc

if i do addn(5) it returns


return lambda x,inc=n: x+inc


<function <lambda> at 0x01D81830>

ok? so what do i write to make it actually do something. and is the
inc=n necessary i cant do x+n?

Here are some notes I have written for our local wiki on lambdas in
python. I hope you will find them illuminating, and I would welcome
any suggestions for improving them.

I have just cut and pasted from our wiki, so the fancy formatting has
been lost.

-----

Python lambdas.

The on-line documentation for python lambdas is not very illuminating.
Here¡¯s my take and my first simple examples.

I would describe a lambda as a parameterised function template. If you
dig, the docs call lambdas anonymous functions not bound to a name.
There is a bit of resemblance to C macros.

Here is a simple lambda that implements an exclusive or:

(Because of the resemblance to C macros, I have been cautious and
written the lambda with lots of parentheses.)

To use this in later code, we define instances of the lambda with
specific function arguments.


We now have two ¡°functions¡±, topping() and sauce() which we can use
later to test flags.
True


So in the definition of the XOR lambda, think of x and y as the
parameters of the function template, and XOR as the function name
placeholder.

By putting in specific objects for the parameters (here the boolean
variables cream and icecream for example), we produce a specific
instance of the lambda, topping() which looks like a function with no
arguments.

If we use different objects for the parameters (say tomato and BBQ)
then we get a different function, sauce.

Here is another simple lambda, (template) to set up three new
functions AddOnly, DeleteOnly, and ReplaceOnly.

#--# Lambda function to check that a flag is only on when the other
two are off. #--#
def TFF(x,y,z) :
return lambda : ( ( x ) and not ( y ) and not ( z ) )

AddOnly = TFF( options.AddAction, options.ReplaceAction,
options.DeleteAction )
DeleteOnly = TFF( options.DeleteAction, options.AddAction,
options.ReplaceAction )
ReplaceOnly = TFF( options.ReplaceAction, options.AddAction,
options.DeleteAction )

if( not (DeleteOnly() or AddOnly() or ReplaceOnly() ) ):
print "Error: Exactly one of [ --add | --replace | --delete ]
allowed. "
parser.print_help()
exit


More advanced lambdas.

The examples above give function instances that have no arguments,
once the parameters of the lambda are chosen.

For a function template with arguments and parameters, we add the
arguments on the 2nd line. Parameters are in the first line.

The Gaussian distribution is exp(-(x-¥ì)©÷/ 2¥ò©÷ ) / ¡î(4 ¥ð¥ò). While we
can think of this as a function of three variables, we normally view
it as a family of functions of a single variable x, parameterised by ¥ì
and ¥ò. Selecting fixed values for ¥ì and ¥ò gives us a single
distribution for x.
... return lambda x : math.exp( - (x-mu)**2 / 2 /sigma**2 ) /
math.sqrt (2 * math.pi *sigma **2 )
...
and here are some instances:

which we later use as
0.0073381331586869951

I recommend defining the instances of the lambda right after the
lambda. If you define it in code far removed from the definition of
the lambda, it looks like an assignment, so comment it.
 
T

Terry Reedy

| I would describe a lambda as a parameterised function template. If you
| dig, the docs call lambdas anonymous functions not bound to a name.

A lambda expression is an abbreviation of a simple def statement:
f = lambda args: expression
def f(args): return expression
have exactly the same effect except that f.func_name will be the less
useful '<lambda>' instead of the more useful 'f'.

| There is a bit of resemblance to C macros.

Macros in C (and, I believe, some places elsewhere) are text-replacement
templates. They are markedly different from function statements. C macros
do not create C functions. Python lambda expression do create Python
function objects. Since C macros are statements, not expressions, and are
introduced by #define, similar to def, one could argue than Python def
statements are more similar to C macros.

| Here is a simple lambda that implements an exclusive or:
|
| >>> def XOR(x,y) :
| >>> return lambda : ( ( x ) and not ( y ) ) or ( not ( x ) and ( y ) )

def XORY(x,y):
def _xory(): x and not y or not x and y
return _xory

has the same effect. Because lambda expressions define functions, not
macros, there is no need for the protective parentheses that macros need.

Here is another simple lambda, (template) to set up three new
| functions AddOnly, DeleteOnly, and ReplaceOnly.
|
| #--# Lambda function to check that a flag is only on when the other
| two are off. #--#
| def TFF(x,y,z) :
| return lambda : ( ( x ) and not ( y ) and not ( z ) )

def TFF(x,y,z):
def _tff(x,y,z): return ( ( x ) and not ( y ) and not ( z ) )
return _tff

Same result (except for a real name in tracebacks), same usage.

| >>> import math
| >>> def Gaussian( mu, sigma ) :
| ... return lambda x : math.exp( - (x-mu)**2 / 2 /sigma**2 ) /
| math.sqrt (2 * math.pi *sigma **2 )

def Gaussian(mu, sigma):
def _gaussian(x): return math.exp( - (x-mu)**2 / 2 /sigma**2 ) /
math.sqrt (2 * math.pi *sigma **2 )
return _gaussian

Again, giving the returned function a name will help a bit if it raises an
exception, which is definitely possible here.

Lambda expressions are an occasional convienience, not a requirement.
Anyone who is confused by what they do should use an equivalent def
instead.

Terry Jan Reedy
 
A

andrej.panjkov

No, no, no, no, no!

Geez. Go easy.
You have got it entirely wrong here. Your XOR function simply returns a
function which gives you the result of xoring the parameters AT THE TIME
WHEN YOU ORIGINALLY CREATED IT. I'm guessing that you had already set
cream and icecream (otherwise the call to XOR would have thrown an
exception) and at leas one was true. Try setting them both False at the
beginning:


False

Ok. I understand this better now. I did say I found the documentation
rather terse on this.
Using a lambda was a completely pointless exercise here, you could have
just returned the result directly:


If I try out a new language, I try to exercise those parts of the
language that are new to me. Now I saw lambdas, an interesting
structure I hadn't seen before. So I tried them out. I get to learn a
little at the same time as scripting. That was the "point". I only
get to optimise my use of a language by trying out various corners of
it.
def TFF(x,y,z) :
return x and not y and not z

AddOnly = TFF( options.AddAction, options.ReplaceAction,
options.DeleteAction )
DeleteOnly = TFF( options.DeleteAction, options.AddAction,
options.ReplaceAction )
ReplaceOnly = TFF( options.ReplaceAction, options.AddAction,
options.DeleteAction )

if not (DeleteOnly or AddOnly or ReplaceOnly):
print "Error: Exactly one of [ --add | --replace | --delete ]
allowed. "
parser.print_help()
exit

which boils down to:

if (options.AddAction + options.ReplaceAction +
options.DeleteAction) != 1:
print "Error: ..."

Indeed, there are many ways this could be done. Some are more
concise, some are more efficient. As I said, I did it the way I did
it to try out lambdas. Your way achieves the result, rather elegantly
I think, but teaches me nothing about using lambdas.

Pardon my tetchiness, but it is a little hard to receive such blunt
and inflexible replies to my posts.

Both the responses offer lambda free alternatives. That's fine, and
given the terse documentation and problems that I had understanding
them, I would agree. So what applications are lambdas suited to? I
think the parameterised function model is one.
What else?
 
G

Gabriel Genellina

En Thu, 08 May 2008 22:57:03 -0300,
No, no, no, no, no! Geez. Go easy.
You have got it entirely wrong here. Your XOR function simply
[...]
Pardon my tetchiness, but it is a little hard to receive such blunt
and inflexible replies to my posts.

Don't take it so seriously. I would have written a reply in the same tone.
Weeds must be uprooted early :)
Both the responses offer lambda free alternatives. That's fine, and
given the terse documentation and problems that I had understanding
them, I would agree. So what applications are lambdas suited to? I
think the parameterised function model is one.
What else?

It should be clear now that lambda is just a shortcut for defining a
normal function using "def", except it has no name, and it can handle
expressions only (no statements).
So you never *need* a lambda. But in a few cases they're useful:

- Most GUIs are event-driven, and let you bind a function (or any other
callable object) to be executed when certain event happens (by example,
when certain button is pressed, or a menu item is selected). Usually an
instance method is used: Button("Total", onclick=self.calculate_total).
Suppose you're developing a calculator; the ten buttons labeled '0' to '9'
should inserte the corresponding digit. To do that, you should write ten
functions insert_digit_0 to insert_digit_9 (and they would be
one-line-functions: insert_digit('0') ... insert_digit('9')). Too boring :(
The usual idiom is something like this:
Button("0", onclick=lambda: self.insert_digit('0'))
Button("5", onclick=lambda: self.insert_digit('5'))

- To write an expression that is to be evaluated lazily (perhaps only if
certain other conditions are met). Older Python versions didn't have a
conditional expression like C's :? ternary operator, and one possible way
to emulate it is this:

def iif(cond, if_true, if_false):
if cond:
return if_true()
else:
return if_false()

iff(x!=2, lambda: 1/(x-2), lambda: 100)

You can't write iff(x!=2, 1/(x-2), 100) because arguments are evaluated
before the function is called, and with x=2 you get an error.
 
L

Lie

No, no, no, no, no!

Geez.  Go easy.


You have got it entirely wrong here. Your XOR function simply returns a
function which gives you the result of xoring the parameters AT THE TIME
WHEN YOU ORIGINALLY CREATED IT. I'm guessing that you had already set
cream and icecream (otherwise the call to XOR would have thrown an
exception) and at leas one was true. Try setting them both False at the
beginning:

Ok. I understand this better now.  I did say I found the documentation
rather terse on this.
Using a lambda was a completely pointless exercise here, you could have
just returned the result directly:

If I try out a new language, I try to exercise those parts of the
language that are new to me.  Now I saw lambdas, an interesting
structure I hadn't seen before. So I tried them out.  I get to learn a
little at the same time as scripting.  That was the "point".  I only
get to optimise my use of a language by trying out various corners of
it.


def TFF(x,y,z) :
  return x and not y and not z
AddOnly = TFF( options.AddAction, options.ReplaceAction,
options.DeleteAction )
DeleteOnly = TFF( options.DeleteAction, options.AddAction,
options.ReplaceAction )
ReplaceOnly = TFF( options.ReplaceAction, options.AddAction,
options.DeleteAction )
if not (DeleteOnly or AddOnly or ReplaceOnly):
  print "Error:  Exactly one of  [ --add | --replace | --delete ]
allowed. "
  parser.print_help()
  exit
which boils down to:
if (options.AddAction + options.ReplaceAction +
        options.DeleteAction) != 1:
    print "Error: ..."

Indeed, there are many ways this could be done.  Some are more
concise, some are more efficient.  As I said, I did it the way I did
it to try out lambdas.  Your way achieves the result, rather elegantly
I think, but teaches me nothing about using lambdas.

Pardon my tetchiness, but it is a little hard to receive such blunt
and inflexible replies to my posts.

Both the responses offer lambda free alternatives.  That's fine, and
given the terse documentation and problems that I had understanding
them, I would agree.  So what applications are lambdas suited to?  I
think the parameterised function model is one.
What else?

Lambda can actually be safely removed from python and no other
features would be missing. It is always possible to create a def
version of any lambda, so lambda is useless. It is just a convenience
for the times where we're just too lazy to invent a name and find a
place to place the def, instead just inlining the function.
 
L

Lie

En Thu, 08 May 2008 22:57:03 -0300,  
<[email protected]> escribió:


No, no, no, no, no! Geez.  Go easy.
You have got it entirely wrong here. Your XOR function simply
[...]
Pardon my tetchiness, but it is a little hard to receive such blunt
and inflexible replies to my posts.

Don't take it so seriously. I would have written a reply in the same tone.  
Weeds must be uprooted early :)
Both the responses offer lambda free alternatives.  That's fine, and
given the terse documentation and problems that I had understanding
them, I would agree.  So what applications are lambdas suited to?  I
think the parameterised function model is one.
What else?

It should be clear now that lambda is just a shortcut for defining a  
normal function using "def", except it has no name, and it can handle  
expressions only (no statements).
So you never *need* a lambda. But in a few cases they're useful:

- Most GUIs are event-driven, and let you bind a function (or any other  
callable object) to be executed when certain event happens (by example,  
when certain button is pressed, or a menu item is selected). Usually an  
instance method is used: Button("Total", onclick=self.calculate_total).  
Suppose you're developing a calculator; the ten buttons labeled '0' to '9'  
should inserte the corresponding digit. To do that, you should write ten  
functions insert_digit_0 to insert_digit_9 (and they would be  
one-line-functions: insert_digit('0') ... insert_digit('9')). Too boring :(
The usual idiom is something like this:
     Button("0", onclick=lambda: self.insert_digit('0'))
     Button("5", onclick=lambda: self.insert_digit('5'))

- To write an expression that is to be evaluated lazily (perhaps only if  
certain other conditions are met). Older Python versions didn't have a  
conditional expression like C's :? ternary operator, and one possible way  
to emulate it is this:

def iif(cond, if_true, if_false):
     if cond:
         return if_true()
     else:
         return if_false()

iff(x!=2, lambda: 1/(x-2), lambda: 100)

You can't write iff(x!=2, 1/(x-2), 100) because arguments are evaluated  
before the function is called, and with x=2 you get an error.

Calling iff would give a NameError. I wonder why... perhaps because
iif is so iffy?
 
A

Arnaud Delobelle

Lie said:
Lambda can actually be safely removed from python and no other
features would be missing. It is always possible to create a def
version of any lambda, so lambda is useless. It is just a convenience
for the times where we're just too lazy to invent a name and find a
place to place the def, instead just inlining the function.

Note that the same thing can be said about generator expressions,
which are nothing more than anonymous, non-reusable, generator
functions. Instead these were _added_ to the language!
 
I

inhahe

Both the responses offer lambda free alternatives. That's fine, and
given the terse documentation and problems that I had understanding
them, I would agree. So what applications are lambdas suited to? I
think the parameterised function model is one.
What else?

i've hardly ever used lambdas since map() and filter() were replaced by list
comprehension. two other uses I can think of for it are: using it as a
sorting key (which takes a function and lambdas are perfect for that when a
direct function isn't available. for example, lambda x: x.myName), and I
made an irc bot once that certain events had a list of fuctions that would
be called after that event. it was like being able to dynamically add and
remove event handlers. for example what if you asked the user a question
and you wanted to know for the next input whether it was from that user and
was an answer to that question. sometimes the function to add would be very
simple, so writing a def for it would just be ugly.
 
T

Terry Reedy

|
| > Lambda can actually be safely removed from python and no other
| > features would be missing. It is always possible to create a def
| > version of any lambda, so lambda is useless. It is just a convenience
| > for the times where we're just too lazy to invent a name and find a
| > place to place the def, instead just inlining the function.
|
| Note that the same thing can be said about generator expressions,
| which are nothing more than anonymous, non-reusable, generator
| functions.

Right. So if someone posted on genexp confusion, I would suggest
'write a full generator function'.

| Instead these were _added_ to the language!

As a convenience.
Actually, if one uses more that one for-clause in a generator expression,
there is a potential gotcha in relation to name capture. So if that bites,
the genexp is not so much a convenience, and one might better write
the full function.

tjr
 
A

Arnaud Delobelle

| Note that the same thing can be said about generator expressions,
| which are nothing more than anonymous, non-reusable, generator
| functions.

Right. So if someone posted on genexp confusion, I would suggest
'write a full generator function'.

I was just arguing against arguing for the removal of lambda on the
basis that it doesn't add any functionality to the language!
| Instead these were _added_ to the language!

As a convenience.
Actually, if one uses more that one for-clause in a generator expression,
there is a potential gotcha in relation to name capture. So if that bites,
the genexp is not so much a convenience, and one might better write
the full function.

tjr

Yes, IMHO this is a bug, and I wish I had the time to dive into the
code to see if I can fix it.
 
B

Bruno Desthuilliers

inhahe a écrit :
i've hardly ever used lambdas since map() and filter() were replaced by list
comprehension. two other uses I can think of for it are: using it as a
sorting key (which takes a function and lambdas are perfect for that when a
direct function isn't available. for example, lambda x: x.myName),

import operator
foos.sort(key=operator.attrgetter('myName'))
 
T

Terry Reedy

| [...]
| > | Note that the same thing can be said about generator expressions,
| > | which are nothing more than anonymous, non-reusable, generator
| > | functions.
| >
| > Right. So if someone posted on genexp confusion, I would suggest
| > 'write a full generator function'.
|
| I was just arguing against arguing for the removal of lambda on the
| basis that it doesn't add any functionality to the language!

I sort of understood that ;-)
Like Guido, I am split on keep/remove.
However, I have decided to leave lambda out of my Python-subset
executable-pseudocode algorithm language. I have not decided whether or
not to include genexps.

| > | Instead these were _added_ to the language!
| >
| > As a convenience.
| > Actually, if one uses more that one for-clause in a generator
expression,
| > there is a potential gotcha in relation to name capture. So if that
bites,
| > the genexp is not so much a convenience, and one might better write
| > the full function.

| Yes, IMHO this is a bug, and I wish I had the time to dive into the
| code to see if I can fix it.

If I do include them, I might restrict them to one for-clause because of
that glitch, whose details I keep forgetting.

tjr
|
 
P

Paul McGuire

i've hardly ever used lambdas since map() and filter() were replaced by list
comprehension.  two other uses I can think of for it are: using it as a
sorting key (which takes a function and lambdas are perfect for that when a
direct function isn't available. for example, lambda x: x.myName), and I
made an irc bot once that certain events had a list of fuctions that would
be called after that event.  it was like being able to dynamically add and
remove event handlers.  for example what if you asked the user a question
and you wanted to know for the next input whether it was from that user and
was an answer to that question.  sometimes the function to add would be very
simple, so writing a def for it would just be ugly.

lambda is handy in defining parse actions in pyparsing. Parse actions
are callbacks to be run when an expression within a larger grammar is
matched. A common use for parse actions is to do some sort of text or
type conversion. The simplest parse actions are called using the list
of matched tokens. Here is a subexpression that will convert numeric
strings found in a larger grammar to ints:

integer = Word("0123456789").setParseAction(lambda tokens:
int(tokens[0]) )

Since this returns an actual int, there is no need to junk up the post-
parsing code with calls to int(), float(), etc. for these simple
conversions.

Here is an example parse action that just converts a set of matched
words to title case:

title = OneOrMore(Word(alphas)).setParseAction(lambda tokens: "
".join([ t.title() for t in tokens ]) )
print title.parseString("the sun also rises")[0]

prints:
The Sun Also Rises

This second example is about as complex as I'd like to get in a
lambda, though. Anything more elaborate than that, and I'd go with a
separately defined function.

-- Paul
 
A

Arnaud Delobelle

[...]
lambda is handy in defining parse actions in pyparsing. Parse actions
are callbacks to be run when an expression within a larger grammar is
matched. A common use for parse actions is to do some sort of text or
type conversion. The simplest parse actions are called using the list
of matched tokens. Here is a subexpression that will convert numeric
strings found in a larger grammar to ints:

integer = Word("0123456789").setParseAction(lambda tokens:
int(tokens[0]) )

Could you use it as a decoratore instead?

integer = Word("0123456789")

@integer.setParseAction
def parse_integer(tokens):
return int(tokens[0])

I could make your grammar clearer, because you don't mix it with
processing code... and no need for lambdas!
 

Ask a Question

Want to reply to this thread or ask your own question?

You'll need to choose a username for the site, which only take a couple of moments. After that, you can post your question and our members will help you out.

Ask a Question

Members online

Forum statistics

Threads
473,744
Messages
2,569,479
Members
44,899
Latest member
RodneyMcAu

Latest Threads

Top