Mark Smith

Yes, maybe the Regular Expression approach may be problematic if the HTML is invalid (which was a disclaimer in my article) but otherwise it seemed to work correctly.

I don't think I'm so keen on the Excel based solution though as it seems to have a lot more overhead (i.e you'd have to try and work out where each table started and ended in the resulting file, and then create a named range for each table via the excel object model, which would then also require excel to be installed to the server). Unless of course I'm mistaken and there's another way to utilise excel in this scenario?

How would you do that Dave? I thought you had to have named ranges to query an Excel file via OleDb? If I'm correct then the only way to create these named ranges would be to open the file with the object model hence needing Excel to be installed.

Charles,

I understand that you can query a spreadsheet, that is not the issue (see the comments above which already go into this). The problem I would have had using that approach is that there would have been multiple tables of completely unrelated data contained within "Sheet1". Therefore, you would end up with a DataTable (or equivilant object) with columns of data that bear no relevance from one table to another and no way of knowing when one table finished and another started. For example, the HTML page could have contained this data:

Table 1:

Field1 Field2

1      2

3      4

Table 2:

Name   Country

Mark   England

Dave   France

Using your approach, I would end up with one DataTable which contains two DataColumns. However, the rows within that DataTable would contain the headers from table 2 along with the data from table 2. Table 1's data has no relevance to Table 2's data, yet you would have them both within one DataTable and therefore no way of knowing when the second table's data started.

The only way around this would be to use the named ranges approach as mentioned above, but this brings in the overhead of having Excel installed on the server.

The method I used, means none of this is necessary so with valid HTML, I think this is the best method I've seen so far.