[rfc][icedtea-web] c-properties-miniparser now escape most (all except multiline?) escape sequences

[Jiri Vanek]

On 12/22/2015 09:58 AM, Jacob Wisor wrote:
> On 12/18/2015 at 04:12 PM Jiri Vanek wrote:
>> SSIA
>>
>> how about 1.6 ? Backport?
>
> This code is supposed to work on manifest files in JAR files and property, files right?

It is parsing only selected items from deployment.properties (jvm home and the plugin arguments)
>
>> +void IcedTeaPluginUtilities::escape(std::string& str) {
>> +    std::string result = "";
>> +    for(unsigned int i = 0; i<str.length(); i++) {
>
> I don't think you want to calculate the length() of str on every loop.
should be fixed

>
>> +//        fprintf(stdout, "%s\n", result.c_str());
>> +        bool processed = false;
>> +        char c1 = str[i];
>> +        if (c1 == '\\') {
>> +            if (i<str.length()-1) {
>
> Same goes here for calculating the length of str on every loop.
should be fixed
>
>> +                char c2 = str[i+1];
>> +                if (c2 == '=' || c2 == '\\' || c2 == ':') {
>> +                    result+=c2;
>> +                    i++;
>> +                    processed = true;
>> +                }
>> +                if (c2 == 't') {
>> +                    result+='\t';
>> +                    i++;
>> +                    processed = true;
>> +                }
>> +                if (c2 == 'n') {
>> +                    result+='\n';
>> +                    i++;
>> +                    processed = true;
>> +                }
>> +                if (c2 == 'r') {
>> +                    result+='\r';
>> +                    i++;
>> +                    processed = true;
>> +                }
>
> I suppose a cascading "if else" structure would be more efficient here.

I had this in first (not published version). It maybe is 0.0000000001% faster but it really do not 
look nicely.

I would much rather stay with this style.
>
>> +
>> +            }
>> +        }
>> +    if (!processed){
>> +        result+=c1;
>> +    }
>> +    }
>> +   str=result;
>> +}
> For one, please fix the formatting and remove trailing white spaces at the end of lines.
>
Should be fixed.

> Second, is this really a good way to do escape sequence processing in C/C++? I'd say that regex

As far as I'm able to judge (with local C gurus)... yes...

As for regexes. I'm not best C man (crap :D)  so I was consulting with others. And .. before I 
consulted, I went with regexes. The result was terrible 9 Tired a *lot*) and I was strongly 
discouraged later.

Then I swapped to this, realised taht this really is it is much much cleaner and better solution.

J.
> substitution and selection is the way to go here. Yes, I am aware that the standard C/C++ libraries
> do not support regex. However, this plug-in implementation is designed and supposed to run on POSIX
> systems, which does support regex since eons. But, before you rush off and start thinking that Perl
> or Python regex implementations/libraries was meant here or they may be a great idea: no they are
> not. Implementing this should not require a huge additional dependency, even if this code uses only
> a small portion of that dependency and possibly has a shorter launching time than a full Java VM.
> regex.h, regcomp(), regexec(), regerror(), and regfree() should be your friends.
>
> With regards to portability, unfortunately, Windows does not natively support regex (because it was
> never intended to be POSIX compliant). However, .NET, Visual Basic and JScript scripting hosts do
> support regex. Of course, .NET is out of the game but the Windows Scripting Host has become a
> default Windows component of the NT family of products since Windows 2000 (and as a service pack for
> Windows NT 4). And, since the Windows plug-in cannot share all native code with the POSIX plug-in
> for other reasons as well, it is only natural to go "the Windows way" where necessary. I do not
> intend to support anything earlier than Windows 2000, so it should be possible and safe to tap into
> the Windows Scripting Host for native regex capability, should regex become a necessity on Windows.
> The Windows Scripting Host has a relatively short launching time despite its long name. ;-)

Well thsi is painful storry. Now I'm happy I'm staying with that iteration.


Thanx!
-------------- next part --------------
A non-text attachment was scrubbed...
Name: CpartEscapeSomeProperties2.patch
Type: text/x-patch
Size: 18770 bytes
Desc: not available
URL: <http://mail.openjdk.java.net/pipermail/distro-pkg-dev/attachments/20151222/d31e8af2/CpartEscapeSomeProperties2-0001.patch>