Digitise ( text )

Returns same psuedo-random number for same text input

Average rating: 4.5 (32 votes) Log in to vote

John Bochan
private
private

Sample input:

Digitise ( "Mary had a little Lamb" )

Sample output:

.113385492832828

Function definition: (Copy & paste into FileMaker's Edit Custom Function window)

/*
Author: John Bochan, 2010
returns the same psuedo-random number for the same text input
requires inbuilt 'Code' function in FMP10 or greater
Input:    any text phrase  (text length is limited by Code function restriction, see FMP documentation)
Output: 0 ≤ real number < 1  (there is no guarantee that this function generates a unique number for each unique text input)
*/
Let (
     [ // convert the input text to a number
       f1 = Substitute ( Code ( text ) ; [ "00" ; "" ] ; [ "0" ; "*" ] ) & 1 ; // create an expression of several numbers multiplied together
       f2 = Replace ( f1 ; Position ( f1 ; "*" ; 1 ; PatternCount ( f1 ; "*" ) / 2 ) ; 1 ; "1/" ) ; // split expression into 2 parts, divide LHS by RHS
       h  = Evaluate ( f2 ) // convert expression into a number
     ] ;
    Mod ( Ln ( h ) ; 1 ) // return the mantissa of the logarithm of this number
)
/*
There is no magic in the text-to-number formula used above, define your own if you like; the magic is in the 'Ln' function (an old trick I used back in the 1970's).
*/

Returns the same pseudo-random number ( 0 <= number < 1 ) for the same input text

Cannot guarantee returned number is unique for each unique text input. In most practical cases it can be considered unique.

Example 1:

How random is your list of sample names? (quick and dirty method)

For N names the Average value of all 'digitised' names should be about 0.5
Standard Deviation of values approx = (gamma/2) where gamma is the Euler-Mascheroni constant
If we sort the values and calculate a difference value between two consecutive values, the Average and Standard Deviation of these differences is approx 1/N

Using Firstname and Lastname data from Brian Dunnings great Free Sample Data file '50000.zip'
- create a new database file and import Firstname and Lastname fields from the sample data file.
- for each record create a number calculation field for the value: Digitise ( FirstName & " " & LastName )
- this list contains 4 duplicated names (FirstName & " " & LastName) resulting in 4 duplicated values
- duplicates can be found by calculated value not name; (calculated field for 'FirstName & " " & LastName' not created)
- The Average of all 50,000 values = 0.500883 ( Average 0.500000)
- The Standard Deviation of these values is 0.288149 (gamma/2 = 0.288608)
- the Average of the difference between ordered consecutive values is 0.000020 (1/N = 0.000020)
- the Standard Deviation of the difference between ordered consecutive values is 0.000020 (1/N = 0.000020 )
- If all calculated values are unique, only need the first [ Floor( 2*Log10(N)+1 ) ] decimal places to maintain uniqueness
- Repeat above using 'Random' instead of 'Digitise ( FirstName & " " & LastName )'; similar statisical results but you cannot find duplicate names by value

Example 2:

Generate a recoverable 4 digit pin number with leading zeroes from several text fields, eg A, B and C
pin = Let ( e = 4 ; Right ( Int ( ( Digitise ( List ( A ; B ; C ) ) + 1 ) * 10^e ) ; e ) )

Comments

		Nick Stockbridge, Transmedia Communications Ltd Jan 22, 2011
I have found this very useful for adding on a string of numbers to a URL. I needed it to slightly scramble a pattern which could have otherwise been predicted in an unwanted way. So thanks, just what I needed!

Note: these functions are not guaranteed or supported by BrianDunning.com. Please contact the individual developer with any questions or problems.

Digitise ( text )

Comments

Support this website.

Support this website.