| Testing validity of numerical content of a string [message #51744] |
Wed, 06 December 2006 17:29  |
Y.T.
Messages: 25
Registered: December 2004
|
Junior Member |
|
|
Maybe this is a stupid question, but is there a way in IDL to tell
whether a string contains a valid representation of a particular data
type? For example an integer? I can do something like
test = fix(str)
which will return zero (and throw an error message) if str is, say,
"hello". But is there a way to tell whether type conversion of a string
will work if attempted?
The context is a bunch of (ascii formatted) data that's been
transmitted through various channels and could conceivably be broken on
a bit-level. So the first sanity check I'd like to perform is to test
whether the received data still "looks like a number" (and then in the
second step I'll perform a couple sanity checks on whether the number
is in a valid range and such).
I am perfectly able and willing to write my own parser that looks at
the individual aspects of the string step-by-step (is the string empty?
Is the first character a digit? etc etc) but parsing for something like
a double (that could be '-.3e-4' or such) would be a major pain.
(I tried "execute" but by itself it just throws the same errors and at
least 6.3/Win it appears that the "quiet" stuff fails:
IDL> test = execute(s)
Type conversion error: Unable to convert given STRING to Double.
Detected at: $MAIN$ 1 C:\Documents and
Settings\idl\test.pro
IDL> test = execute(s,/QuietExecution)
test = execute(s,/QuietExecution)
^
Keyword parameters not allowed in call.
Dunno. )
Anybody have something clever at hand?
|
|
|
|
| Re: Testing validity of numerical content of a string [message #51821 is a reply to message #51744] |
Thu, 07 December 2006 10:18  |
Brian Larsen
Messages: 270
Registered: June 2006
|
Senior Member |
|
|
Excellent solution. I haven't played with on_ioerror but certainly
will now.
Funny how you get stuck in a rut of how you normally do things even if
its not the best, in fact probably never the best way :)
Brian
mgalloy@gmail.com wrote:
> Sven Geier wrote:
>> In a pinch one might just force the conversion and catch any resulting
>> error-messages with ON_ERROR -- that's ugly, but might be a lot easier than
>> trying to parse strings...
>
> So here's some code that does that:
>
> function validate_numeric, value
> compile_opt strictarr
>
> on_ioerror, badValue
>
> f = float(value)
> return, 1B
>
> badValue: return, 0B
> end
>
> The odd thing about this is what IDL will do without complaining. IDL
> will do this without any error:
>
> IDL> help, float('0.0hello5')
> <Expression> FLOAT = 0.00000
>
> but draws the line at:
>
> IDL> help, float('hello5')
> % Type conversion error: Unable to convert given STRING to Float.
> % Detected at: $MAIN$
> <Expression> FLOAT = 0.00000
>
> Mike
> --
> www.michaelgalloy.com
|
|
|
|
| Re: Testing validity of numerical content of a string [message #51822 is a reply to message #51744] |
Thu, 07 December 2006 10:08 |
Bob[3]
Messages: 60
Registered: December 2006
|
Member |
|
|
On Dec 7, 12:43 pm, "mgal...@gmail.com" <mgal...@gmail.com> wrote:
> The odd thing about this is what IDL will do without complaining. IDL
> will do this without any error:
>
> IDL> help, float('0.0hello5')
> <Expression> FLOAT = 0.00000
>
> but draws the line at:
>
> IDL> help, float('hello5')
> % Type conversion error: Unable to convert given STRING to Float.
> % Detected at: $MAIN$
> <Expression> FLOAT = 0.00000
Looks like Sven's regex is close to what IDL is doing. (except that
both '+' and '-' are accepted and any exponent can have any number of
digits, try e.g. help, float(' +5e00001hello')
|
|
|
|
| Re: Testing validity of numerical content of a string [message #51824 is a reply to message #51744] |
Thu, 07 December 2006 09:43 |
Michael Galloy
Messages: 1114
Registered: April 2006
|
Senior Member |
|
|
Sven Geier wrote:
> In a pinch one might just force the conversion and catch any resulting
> error-messages with ON_ERROR -- that's ugly, but might be a lot easier than
> trying to parse strings...
So here's some code that does that:
function validate_numeric, value
compile_opt strictarr
on_ioerror, badValue
f = float(value)
return, 1B
badValue: return, 0B
end
The odd thing about this is what IDL will do without complaining. IDL
will do this without any error:
IDL> help, float('0.0hello5')
<Expression> FLOAT = 0.00000
but draws the line at:
IDL> help, float('hello5')
% Type conversion error: Unable to convert given STRING to Float.
% Detected at: $MAIN$
<Expression> FLOAT = 0.00000
Mike
--
www.michaelgalloy.com
|
|
|
|
| Re: Testing validity of numerical content of a string [message #51834 is a reply to message #51744] |
Wed, 06 December 2006 22:21 |
Sven Geier
Messages: 17
Registered: July 2002
|
Junior Member |
|
|
Brian Larsen wrote:
> I just read you post a little closer and the numbers you have are more
> complicated but not too bad.
>
> IDL> print, stregex('hello', '[a-df-zA-DF-Z]', /boolean)
> 1
> IDL> print, stregex('3.14159', '[a-df-zA-DF-Z]', /boolean)
> 0
> IDL> print, stregex('3.14159e3', '[a-df-zA-DF-Z]', /boolean)
> 0
>
> To make it work perfectly you would need to know a little more info
> about what could be there but if's based on these regular expressions
> can tell the difference between 'hello' and '-.3e-4' which can then
> build valid_float.pro.
I kinda mostly agree with the regexp direction, but I'd recommend looking
for a lot more than just "what characters are in the string". Order counts,
for example: -1.e4 -1e.4 1e.-4 1-e4. all contain the same characters, but
not all are well-formed doubles.
(And if the OP manages to get bit-errors that turn "3.141" into "hello",
then I'd like to meet those bit-errors... ;)
I'm somewhat rusty on this, but the full regexp for a double would probably
be something something like
*-?[0-9]*\.?[0-9]*([de]-?[0-9]{1,2})?
i.e. zero or more spaces, then zero or one minus sign, then zero or more
digits, then zero or one literal period followed by zero or more additional
digits followed by zero or one instance of (the letter 'e' or 'd' followed
by zero or one minus sign followed by one or two digits).
And this probably doesn't cover all of it...
I always found IDLs implementation of the regexp library to be rather slow;
so this might not be practical for a large amount of data anyways.
In a pinch one might just force the conversion and catch any resulting
error-messages with ON_ERROR -- that's ugly, but might be a lot easier than
trying to parse strings...
- S
--
http://www.sgeier.net
My real email address does not contain any "Z"s.
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
