Creating a Formula Parser that can work with Dates, Strings, and Numbers
Posted on 2004-08-16
i am puzzled with a problem that came up in my application. I have built an application that shows a list of Data Items (lets assume it would be Tasks for a project) in an Excel style manner, in a HTML form. The display which tasks are actually shown to the user is configurable by the user by the means of a formula. There currently exists code to enter a simple formula like "status=open" - where the application will filter the list of tasks based on the value contained in the status field of the individual tasks. In the example, it would only show tasks that have the value "open" set.
I would like to make this system much more flexible now. I support a few different types of data, most noteworthy dd.mm.yyyy date formats, integers, and Strings. What i would like to create is a system that allows me to create arbitray formulas that will decide whether an Item will be shown or not. For this, I though about creating something like this:
1) the system would still decide on the basis of field<operator>condition --- where operator can be one of ==,~=,!=,<,>,<=,>=
2) first, the system should look for so-called macros -- basically placeholders for values provided by the system --- e.g. today() would evaluate to 16.08.2004 or title() would yield the current items title --- lets assume there is a working set of these macros available.
3) the condition field should be able to notice when it finds a date, and apply further math to the field based on date processing logic -- for example: enddate<22.12.2004+2 would show all items that have end dates smaller than 24.12.2004
4) if no dates are found the system should try to create an integer result out of the formula , e.g. "completeinpercent>32+18" would match all items with the completeinprecent field set to 51 and higher
5) if it all fails, concatenate the strings and compare to the field value -- e.g. title=="Hello"+3+"World" would match the "Hello3World" titled items only
6)Finally, if all of this fails, it should error out.
The tricky bit on this for me is how to implement the parser in a flexible enough way to handle the comparisons of different kinds of data, while being open to formulas with several elements, maybe only noticing the final return data type in the last step. For example, the formula 2+3+"bar" only knows its going to return "23bar" on encounter of the last string as per the rules above.
I am REALLY curious if someone can help me find a good, flexible solution path for this. I will not need a fully fledged implementation, only a hint how this is to be done properly and open to future expansion for more different types of data.