Regular expression to split one big html table to several tables of 5 rows

Question

I'm trying to sort out with regExps, so I met a problem on my way: The thing is I have some random HTML file with plain text and only one table. Text can be before and after the table, table doesn't include <thead><tbody><tfoot> rowspan and so on. So I need to split this table into several tables with 5 rows each and 5 or less the last one, with repeating first string of the original table in each table. So for example:

<table>
  <tr>
   <td>A</td><td>B</td>
  </tr>
  <tr>
   <td>A1</td><td>B1</td>
  </tr>
  <tr>
   <td>C</td><td>D</td>
  </tr>
  <tr>
   <td>E</td><td>F</td>
  </tr>
  <tr>
   <td>E1</td><td>F1</td>
  </tr>
  <tr>
   <td>E2</td><td>F2</td>
  </tr>
  <tr>
   <td>E3</td><td>F3</td>
  </tr>
  <tr>
   <td>E4</td><td>F4</td>
  </tr>
</table>

Should become:

<table>
  <tr>
   <td>A</td><td>B</td><--!!!(not needed to be in code)-->
  </tr>
  <tr>
   <td>A1</td><td>B1</td>
  </tr>
  <tr>
   <td>C</td><td>D</td>
  </tr>
  <tr>
   <td>E</td><td>F</td>
  </tr>
  <tr>
   <td>E1</td><td>F1</td>
  </tr>
</table>
<table>
  <tr>
   <td>A</td><td>B</td><--!!!(not needed to be in code)-->
  </tr>
  <tr>
   <td>E2</td><td>F2</td>
  </tr>
  <tr>
   <td>E3</td><td>F3</td>
  </tr>
  <tr>
   <td>E4</td><td>F4</td>
  </tr>
</table>

This stuff I need to be done using PCRE in PHP including massives of templates and changes. So I have problems with realisation. For now I can find the first row like this <table>\s*?(<tr>(?:\s|.)*?<\/tr>) and 4 going one-by-one rows (<tr>(?:\s|.)*?<\/tr>\s*){1,4} but I cannot get how should I find all the occurrences of the second template so I can use them later on and how to stop searching if there is </table> table closing tag. So please help

EDIT

question has been answered so the next level of it to add in original table tags <thead><tbody><tfoot>. In output tables the structure of original table should be reconstructed, so I mean if the first row of original table was part of <thead> tag it should be in <thead> is all output tables.

Answer 1

You can achieve this by performing a loop where each iteration will add the next "table break" with preg_replace (but see disclaimer at the end). The proposed regular expression will find the following groups:

• The last occurrence of the <table> tag together with the first row that follows it, or, if there is a thead and/or tbody tag, up until the closing </thead> tag, including the opening <tbody> tag if there is one.
• The 4 next rows that follow that. There must be 4 of them.

It then also looks ahead to ensure that there is at least one more row present.

With that information a single "table break" can be injected into the HTML string.

If the table has a tfooter part (which should then also be repeated in each partition of the table), we will not yet have that information, as it occurs at the very end of the input. For that reason a separate parsing is necessary, before the loop starts, to extract the footer.

This is the code assuming the input is in the variable $html:

// Extract the footer part (if there is one) and closing table tag
preg_match("#(\s*(</tbody|<tfooter).*?)?</table>#s", $html, $tableEnd);
$tableEnd = $tableEnd[0];

// Add a table break in each iteration as long as the last partition has more than 4 rows:
while (true) {
    $res = preg_replace("#(<table(?!.*<table).*?/tr>(?:.*?/thead>)?(?:.*?<tbody>)?)((?:.*?/tr>(?=\s*<tr)){4})#s", 
                        "$1$2$tableEnd\n$1", $html);
    if (strlen($res) === strlen($html)) break;
    $html = $res;
}

echo $res;

See it run on eval.in.

Explanation of the main regex

Here are a few highlights in the main regex:

• #: I used this as regex delimiter instead of / to avoid the need to escape / inside the regex itself. If you need to use / as delimiter then escape each / as \\/: one backslash is for the regex and one more to escape that backslash in the context of the string literal.

• (?!.*<table): makes sure that there is no other <table> tag following the one we are about to match. It is a negative look ahead.

• ((?:.*?/tr>(?=\s*<tr)){4}): grabs 4 rows, and with positive look ahead ((?= )) requires that each row is immediately followed by another row. The (?: ) pattern does not make a capture group, but the outer parentheses do create one.

The replacement

If the replacement would just inject again the 2 captured groups (i.e. $1$2), then nothing would have changed. The additional $tableEnd\n$1 will close the table (with the footer) and start a next one by reusing the first capture group. That will have the opening <table> tag with the first row and/or table header included.

Disclaimer

Although the above may work in many cases, it is quite possible to break it, as regular expressions are not the ideal way to parse/interpret HTML. You should really use a DOM api for that, and PHP has one: DOMDocument.