Regexp and Pattern.class

[roger_varley]

Hi

I've got an application (over which I have no control) that presents
its data as a single string. The data contains ' (single quote)
characters that denote end of line. However, the data can also
legitimately contain the ' character, so the generating program escapes
any embedded ' characters with ? (Question mark). (Its a Tradacomms
formatted EDI file if anyone is interested).

How/Can I phrase the regexp parameter to the Pattern.split() method to
split the string back into the original lines. Once I've cracked this,
the + and : characters used to split each line into groups and
individual fields should be easy

Or am I going to have to hand-roll this by reading the string a
character at a time?

Regards
Roger

 

[Tilman Bohn]

In message <[email protected]>,

Hi

I've got an application (over which I have no control) that presents
its data as a single string. The data contains ' (single quote)
characters that denote end of line. However, the data can also
legitimately contain the ' character, so the generating program escapes
any embedded ' characters with ? (Question mark). (Its a Tradacomms
formatted EDI file if anyone is interested).

First question: Can a question mark followed by an apostrophe be
legal application data? If so, how is the question mark or the
complete sequence escaped?

For now I'll assume the sequence ?' can never occur legally in
the application data.

How/Can I phrase the regexp parameter to the Pattern.split() method to
split the string back into the original lines.

Under the above assumption you would split either on "(?<!\\?)'"
or on "(?<=[^?])'", according to taste. The look-behind assertions
are needed so the last character of each line isn't cut off.

Once I've cracked this,
the + and : characters used to split each line into groups and
individual fields should be easy

So no help needed there then. Ok. ;-)

Or am I going to have to hand-roll this by reading the string a
character at a time?

Nope. The above should work.

 

[roger_varley]

First question: Can a question mark followed by an apostrophe be
legal application data? If so, how is the question mark or the
complete sequence escaped?

I've never seen that combination in <mumble> years of handling
Tradacomms EDI files so I've had to actually go and test it. The
generating program throws out ???' where the sequence ?' occurs.

For now I'll assume the sequence ?' can never occur legally in
the application data.

Thanks for your help.

Regards
Roger

 

[klynn47]

Sometimes I find it easier to use the Unicode representation of certain
characters.

 

[Tilman Bohn]

In message <[email protected]>,
(e-mail address removed) wrote on 17 Dec 2004 09:24:20 -0800:

[...]

I've never seen that combination in <mumble> years of handling
Tradacomms EDI files so I've had to actually go and test it. The
generating program throws out ???' where the sequence ?' occurs.

Interesting. Ok, in this case the pattern I gave you won't work
correctly. Before you can find the correct one you'll need to try
what happens for a) ??' and b) ???'.

 

[klynn47]

Sometimes I find it easier to use the Unicode representation of some
characters.

 

[klynn47]

Sometimes I find it easier to use the Unicode representation of certain
characters.

 

[Tilman Bohn]

In message <[email protected]>,
(e-mail address removed) wrote on 17 Dec 2004 08:19:21 -0800:

[...]

How/Can I phrase the regexp parameter to the Pattern.split() method to
split the string back into the original lines.

BTW, that's backwards. The regexp gets passed to Pattern.compile()
first, then your input is the parameter to the split() method executed
on the resulting Pattern object.

 

[roger_varley]

Hi Tilman

??' in the input results in ?????' in the output file and ???' in the
input file results in ???????' in the output.

Regards
Roger