[RESOLVED] ereg_replace question

mrbaseball34

I am utterly terrible at regex's and I need the clause to replace all chars EXCEPT

A-Z
a-z
0-9
- 
\ 
/ 
: 
[ 
]
(
)
#

I have this to begin, just don't know how to add the other chars:

"[^A-Za-z0-9]"

Help is definitely appreciated.

bradgrafelman

First of all, the ereg_*() functions have been deprecated for some time now; you should move to the [man]PCRE[/man] family of regexp functions (eg. [man]preg_replace/man).

mrbaseball34 wrote:
just don't know how to add the other chars

Just add the characters to the list. Don't forget to escape any characters that might have special meaning in regular expression syntax, e.g. the hyphen (unless it's used as the last character in the character class), the square brackets, and whichever character you use as the delimiter (commonly '/', but you can use something else if you'd like, e.g. $).

It goes without saying, of course, that you also must escape any characters that have special meaning as far as PHP is concerned, such as the backslash.

EDIT: Also note that "A-Za-z" could be replaced with simply "a-z" (or "A-Z", whichever you prefer) if you add the 'i' modifier after the closing delimiter (making the pattern case-insensitive).

mrbaseball34

The reason behind this is that we allow people to upload CSV files to our server and they are parsed into a database.

Some users upload, whether accidentally or not, it looks like the binary XLS files and the parsing routing we are using, fgetscv doesn't skip the records that have binary chars.

We need a routine to remove those chars from our table after they are uploaded.

Is there some way to ignore these files or records during the import process?

Or do you have another way??????

bradgrafelman

mrbaseball34 wrote:
Or do you have another way??????

There might be a better/different way to do it, but we'd probably have to see some sample data that you're trying to sanitize/filter.

mrbaseball34

Is there a way to determine if a file is binary or text?

bradgrafelman

I suppose you could search for non-printable characters and hope that covers you... otherwise, no, since text is still binary data (it just so happens that there's a conversion available, e.g. ASCII, that translates "0100100001101001" to "Hi").