regex for html parsing (in c#)

Question

Hello,

I'm trying to parse a html page and extract 2 values from a table row. The html for the table row is as follows: -

<tr>
<td title="Associated temperature in (ºC)" class="TABLEDATACELL" nowrap="nowrap" align="Left" colspan="1" rowspan="1">Max Temperature (ºC)</td>
<td class="TABLEDATACELLNOTT" nowrap="nowrap" align="Center" colspan="1" rowspan="1">6</td>
<td class="TABLEDATACELLNOTT" nowrap="nowrap" align="Center" colspan="1" rowspan="1"> 13:41:30</td>
</tr>

and the expression I have at the moment is:

<tr>[\s]<td[^<]+?>Max Temperature[\w\s]*</td>[\s]
<td[^<]+?>(?<value>([\d]+))</td>[\s]
<td[^<]+?>(?<time>([\d\:]+))</td>[\s]</tr>

However I don't seem to be able to extract any matches. Could anyone point me in the right direction, thanks.

Answer 1

When you write <td[^<]+?> I guess you really mean <td[^>]*>

That is "opening brace, td, maybe stuff other than closing brace..."

Answer 2

<tr>[\s]<td[^<]+?>Max Temperature[\w\s]*</td>[\s]

Not looked at it all yet, but that [^<] probably needs to be [^>] as you're trying to match all non-> until the > that's before Max temperature.

Answer 3

Try

<tr>\s*
<td[^>]*>.*?</td>\s*
<td[^>]*>\s*(?<value>\d+)\s*</td>\s*
<td[^>]*>\s*(?<time>\d{2}:\d{2}:\d{2})\s*</td>\s*
</tr>\s*

Answer 4

The " (ºC)" before the closing td was matched against:

<tr>[\s]<td[^<]+?>Max Temperature[^<]*</td>[\s]

Is that \w a word-boundary? I think that it gets a little tricky there, I'd use a more general approach.

And on the third line, there is one whitespace after the td tag, is that accounted for?

<td[^<]+?>[\s]?(?<time>([\d\:]+))</td>[\s]</tr>

Answer 5

Parsing HTML reliably using regexp is known to be notoriously difficult.

I think I would be looking for a HTML parsing library, or a "screen scraping" library ;)

If the HTML comes from an unreliable source, you have to be extra careful to handle malicious HTML syntax well. Bad HTML handling is a major source of security attacks.

Answer 6

I use http://www.regexbuddy.com/ for such controls. So far I tested @sgehrig's suggestion is correct