Hi Richard and Roelof,
thanks for your comprehensive answer. I brought up Regex only to point out
alternative solutions. Another one is the following using transducers,
where Tee works like the tee command from the command line.
IsbnValidator>>isValidIsbn: aString
| length countChars separators getSeparators lastChar getLastChar
filterDigits computeCheckSum checkSum |
"Count number of characters"
length := 0.
countChars := [:count :char | length := length + 1] init: length.
"Get non-digit characters"
separators := Set new
getSeparators := separators <~ #isDigit remove.
"Get last character"
lastChar := nil.
getLastChar := [:prev :char | lastChar := char ] init: lastChar.
"Get digits"
filterDigits := #isDigit filter.
"Calculate check sum"
computeCheckSum := ([:sum :index :digit | sum + index * digit value]
init: 0) completing: #\\ .
"Compute"
checkSum := aString
transduce: (Tee to: countChars) * (Tee to: getSeparators) * (Tee to:
getLastChar) * filterDigits
reduce: computeCheckSum
init: 0.
"Check validity"
^((length = 10 or: [length = 13])
and: [separators = Set with: $-])
and: [checkSum = (lastChar = $X ifTrue: [10] ifFalse: [lastChar
value])]
Kind regards,
Steffen
Am .09.2020, 08:30 Uhr, schrieb Roelof Wobben via Pharo-users
<pharo-users@lists.pharo.org>:
Op 8-9-2020 om 04:22 schreef Richard O'Keefe:
There are two quite different questions.
(1) Where may dashes occur in a real ISBN-10?
(2) What does Exercism require in the specification and check in the
test cases?
For (1) the rules are
Each ISBN consists of 5 elements with each section being separated by
spaces or hyphens. Three of the five elements may be of >>varying
length:
Prefix element – currently this can only be either 978 or 979. It is
always 3 digits in lengthRegistration group element – this identifies
the particular country, geographical region, or language area
participating >>in the ISBN system. This element may be between 1 and 5
digits in lengthRegistrant element - this identifies the particular
publisher or imprint. This may be up to 7 digits in lengthPublication
element – this identifies the particular edition and format of a
specific title. This may be up to 6 digits >>in lengthCheck digit –
this is always the final single digit that mathematically validates the
rest of the number. It is >>calculated using a Modulus 10 system with
alternate weights of 1 and 3.An ISBN-10 does not have the three-digit
prefix. So we have
[0-9]{1,5} -- prefix
[0-9]{1,7} -- registrant
[0-9]{1,6} -- publication
[0-9X]
-- check digit
As an examplw, "Standard C++ IOStreams and Locales" by Langer & Kreft
has
ISBN-10 0-201-18395-1
ISBN-13 9780201183955
so I shall assume the separators are optional.
/^[0-9]{1,5}[- ]?[0-9]{1,7}[- ]?[0-9]{1,6}[- ]?[0-9X]$/
Of course the elements cannot all have their maximum length at the same
time. In AWK I would write
x = a_putative_ISBN_10
y = x
gsub(/[- ]+/, "", y)
if (x ~ /^[0-9]{1,5}[- ]?[0-9]{1,7}[- ]?[0-9]{1,6}[- ]?[0-9X]$/ \
&& y ~ /^[0-9]{9,9}[0-9X]$/ \
) {
x *might* be valid, we still need to check the checksum
}
For (2), there appear to be no restrictions on where dashes may occur
or how many: "These may be communicated with or without hyphens".
Exercism doesn't allow spaces.
Regular expressions are elegant in their own way, BUT for this problem
they are (a) excessive, (b) inefficient, and (c) insufficient.
digit count := 0.
check sum := 0.
for each character c of the string
if c is not a hyphen then
if c is a digit then
digit value := c's value as a digit
else if c is X and digit count = 9 then
digit value := 10
else
return false.
digit count := digit count + 1.
if digit count > 10 then return false.
check sum := (11 - digit count) * digit value + check sum.
return check sum mod 11 is zero.
Part of the insight here is "don't DO it, just PRETEND you did."
That is, instead of copying the string without the hyphens,
just ignore the hyphens as they turn up.
Another part is "if you are only going to use it once, don't store it."
That is, we need a digit's value just once, in the update to check sum,
so we should compute it just before we need it, not store it.
Now the pseudo-code above is classic sequential imperative coding.
Classic functional coding does something like
let no_dashes = filter (/= '-') (explode string) in
length no_dashes = 10 and
let check = last no_dashes in
(is_digit check or check = 'X') and
all is_digit (take 9 no_dashes) and
let xval c = if x = 'X' then 10 else digit_value c in
dot (map xval no_dashes) [10,9..1]) mod 11 = 0
This pseudo-code translates nicely to Smalltalk too.
You might want to add
SequenceableCollection>>
with: other inject: initial into: aBlock
|r|
r := initial.
self with: other do: [:x :y |
r := aBlock value: r value: x value: y].
^r
dot: other
^self with: other inject: 0 into: [:acc :x :y | x*y + acc]
(These methods are so obvious that it would be absurd to claim
any intellectual property rights to them.)
I also have "fusion" methods like
SequenceableCollection>>
from: start to: finish allSatisfy: testBlock
self from: start to: finish do: [:each |
(aBlock value: each) ifFalse: [^false]].
^true
so that ((seq copyFrom: a to: z) allSatisfy: blk)
can be done as (seq from: a to: z allSatisfy: blk)
without making a copy.
Fusion methods are useful because Smalltall compilers
don't work as hard at eliminating intermediate data
structures as functional language compilers. (Having
other priorities.)
(isdigit (last no_dashes return false if digit count > 10.
Thanks for letting me see this.But still I wonder if this is really the
OOP way and if that function does more then 1 thing.It looks to me that
its iterating trrough the string. Calculating the crc and checking it.I
learned that it is a good thing that a function and a class does only 1
thing.
Roelof
