How to get table structures from a .frm file using PHP? - php

Question

Let I have created a table named after pb00k via phpMyAdmin which SQL is as bellow:
CREATE TABLE `pb00k` (
`k3y` int(255) NOT NULL AUTO_INCREMENT,
`n4m3` text NOT NULL,
`numb3r` text NOT NULL,
PRIMARY KEY (`k3y`)
) ENGINE=InnoDB DEFAULT CHARSET=utf8;
So it generates a pb00k.frm file in mysql/data/ which contains this structure. Now I want to get this structure from this file using PHP.
Is it possible?

Answer 1

Yes it's possible to recover at least part of information. (For the benefit of other readers the poser of the question is already aware that there are easier ways to get the column metadata).
The challenge is that .frm files are not so well documented because any need to decipher them by the general community is pretty rare. Also the format of the files may vary with the operating system.
However, by viewing the files with hexdump or a similar utility you can see partly what is going on. Then you better informed to read the files in a PHP program and decode the raw binary data.
I did this as an exercise some time back, and I was able to recover number of columns, column names and column types.
Below is a sample to show how to extract column names. My .frm was for a table names "stops", but you can substitute your own .frm.
<?php
$fileName = "stops.frm";
// read file into an array of char
//---------------------------------
$handle = fopen($fileName, "rb");
$contents = fread($handle, filesize($fileName));
fclose($handle);
$fileSize=strlen($contents); // save the filesize fot later printing
// locate the column data near the end of the file
//-------------------------------------------------
$index = 6; // location of io_size
$io_size_lo = ord($contents[$index]);
$io_size_hi = ord($contents[$index+1]);
$io_size = $io_size_hi *0x100 + $io_size_lo; // read IO_SIZE
$index = 10; // location of record length
$rec_len_lo = ord($contents[$index]);
$rec_len_hi = ord($contents[$index+1]);
$rec_len = $rec_len_hi * 0x100 + $rec_len_lo; // read rec_length
// this formula uses io_size and rec_length to get to column data
$colIndex = ( ( (($io_size + $rec_len)/$io_size) + 1) * $io_size ) + 258;
$colIndex -= 0x3000; // this is not documented but seems to work!
// find number of columns in the table
//-------------------------------------------------
echo PHP_EOL."Col data at 0x".dechex($colIndex).PHP_EOL;
$numCols = ord($contents[$colIndex]);
//Extract the column names
//--------------------------------------
$colNameIndex = $colIndex+0x50; //0X50 by inspection
echo "Col names at 0x".dechex($colNameIndex).PHP_EOL;
$cols=array();
for ($col = 0; $col < $numCols; $col++){
$nameLen = ord($contents[$colNameIndex++]); // name length is at ist posn
$cols[]['ColumnName']= substr($contents,$colNameIndex,$nameLen-1); // read the name
$colNameIndex+=$nameLen+2; // skip ahead to next name (2 byte gap after \0)
}
print_r($cols);
This should get you started. I will add to this when I have time in the coming days if you think is heading in the right direction.
EDIT. I updated the code so it should work for any .frm file (from Table). For sure there is a free tool to recover mySQL (based on innoDB engine) available at https://github.com/twindb/undrop-for-innodb. Having read through the code and the associated blogs, they are not using the .FRM files for recovery. The same table information is also stored in the innoDB dictionary and they are using this to recover table formats etc.
There is also a way to read the .FRM files content. This described here https://twindb.com/how-to-recover-table-structure-from-frm-files-online/. However, they are using mySQL to read the .frm files and recreating tables from there.
There is also a utility a package of utilities found here https://www.mysql.com/why-mysql/presentations/mysql-utilities/ that contains a .frm reader. This was made by Oracle, who are the only people who know the format of the .frm files! The utility is free so you can download it.
Oracle publish some information on the format of .frm files https://dev.mysql.com/doc/internals/en/frm-file-format.html, but it is both incomplete and incorrect! See this previous Stack question.
https://dba.stackexchange.com/questions/208198/mysql-frm-file-format-how-to-extract-column-info
Now after all that if you still want to try to parse the .frm files yourself for fun or for learning, then you need to be patient and spend time unravelling quite a complicated structure. If you want to keep trying that is OK but send me your .FRM file ( to sand_groper80#hotmail.com) so I can check it out and I will send you some PHP code in a few days that will extract some additional information like datatype and display sizes.