Don't match full string regex

MBartlett21

How do i not match a full string in regex (I want to extract numbers out of it)

kevin_o_toole_1

If you include a capture group in your regexp e.g. [a-z]^([1-9]+) , the value of that group will be in the resulting "captures" array.

From the match() documentation:

captures {array} : The first element is always the input string s. The following elements are a list of all captured groups

ilya_baran

Adding to Kevin's response: you can always add .* on both sides of a regexp to make it match a substring.  You should also look at REGEX_NUMBER_CAPTURE -- that can help you extract real numbers (not just integers) from a regexp.

Finally, a much more reliable alternative to parsing strings is the lambda-based expression evaluation that Mahir and I used for the parametric curve and surface features: https://cad.onshape.com/documents/578ff8b3e4b0e65410fcfda3/w/d33395f174e5b38f4abd6097/e/0c17de6a800d4aed83de417f

mthiesmeyer

Hi @mbartlett21

Perhaps add an example string that you would like to match and some examples that should not match? Are you saying you would like a regex that matches any string that contains at least one number? Or contains all numbers? Or something else?

MBartlett21

@mthiesmeyer
I would like a function to evaluate a string (like javascript eval()).
Example string: "x^2 + y^0.5".

I was wanting to use something like this to get the numbers in the string

kevin_o_toole_1

If you include a capture group in your regexp e.g. [a-z]^([1-9]+) , the value of that group will be in the resulting "captures" array.

From the match() documentation:

captures {array} : The first element is always the input string s. The following elements are a list of all captured groups

ilya_baran

Adding to Kevin's response: you can always add .* on both sides of a regexp to make it match a substring.  You should also look at REGEX_NUMBER_CAPTURE -- that can help you extract real numbers (not just integers) from a regexp.

Finally, a much more reliable alternative to parsing strings is the lambda-based expression evaluation that Mahir and I used for the parametric curve and surface features: https://cad.onshape.com/documents/578ff8b3e4b0e65410fcfda3/w/d33395f174e5b38f4abd6097/e/0c17de6a800d4aed83de417f

MBartlett21

@ilya_baran
How does the Expression type work?

ilya_baran

It's all defined in that feature studio.  Basically, the idea is that an Expression is just a lambda function, say of one variable, t.  The operators are overloaded for Expression types so that when you, for example, add two Expressions, the result is a function that evaluates the two functions and returns their sum.  Constants are just turned into functions that return that constant, and t_ is the identity function.  So when you write t_ + 3, you get a function that returns 3 plus its argument.  More complicated expressions also just turn into functions that evaluate them.

MBartlett21

@ilya_baran
Is it faster evaluating through a few functions to get the value, or would functions that returned reverse polish notation be faster?

ilya_baran

Neither will be terribly fast.  I'm not sure which is faster -- you'd have to measure.  I have a vague plan of using operator/function overloading to construct an expression tree in FS and have builtin functions compile it using LLVM, but that's far out.

MBartlett21

What is LLVM?

owen_sparks

mbartlett21 said:

What is LLVM?

https://en.wikipedia.org/wiki/LLVM

MBartlett21

Ilya, would there be any way to have an input like

annotation { "Name" : "Input", "Variables" : ["var1","var1"]}
definition.input is Expression;

MBartlett21

Here is my evalExpression script now (it first splits the tokens, turns that to Reverse Polish Notation, then evaluates it. It also allows contxts and lists of variables to be input, like x, y, etc.):
https://cad.onshape.com/documents/429fd7dda5f0fdb93833a34a