Split each expression in String

[Vani Kumar]

how to split each expression in the below String (string can have any combination of criteria)

e.g., name eq 'alex' or role eq 'admin' and salary gt '3000'

Need to split it as below
name eq 'alex'
role eq 'admin'
salary gt '3000'

(I have a RESTful api that might have different combination of unexpected criteria in the uri)

Do i need to write any regular expression for this?

[Campbell Ritchie]

You could try splitting on "and|or|implies|iff|xor" or similar

[fred rosenberger]

an example is great, but it doesn't make a spec.

Can you split on spaces, then recombine three and discard the fourth?

[Vani Kumar]

That works, but i can have different number of combinations in the string like below

example 1:
"role eq 'admin' and department eq 'science'"

example 2:
"name eq 'vick' and role eq 'admin' and salary gt '3000'" or status eq 'active'

example 3:
"(address/zipcode eq '11122' or address/zipcode eq '22233') and (name/lastname eq 'xang' or name/firstname eq 'john') and status eq 'active' and contains(familyName, 'smith')";

[Vani Kumar]

Actually, to be more specific, i need to get like below e.g., for example 3

(address/zipcode eq '11122' or address/zipcode eq '22233')
and
(name/lastname eq 'xang' or name/firstname eq 'john')
and
status eq 'active'
and
contains(familyName, 'smith')

and then again i can split it like below

address/zipcode eq '11122'
or
address/zipcode eq '22233'
and
name/lastname eq 'xang'
or
name/firstname eq 'john'
and
status eq 'active'
and
contains(familyName, 'smith')

[Campbell Ritchie]

What are the associativities and precedences of the different operators, and, or, etc.

[Henry Wong]

Based on the examples, there are a few concerns -- concerns that make the parsing complicated.

Vani Kumar wrote:
example 1:
"role eq 'admin' and department eq 'science'"

First, the identifier may have "and" or "or" embedded in it. For example, what about "neighborhood eq 'kew gardens'". There is actually an "or" in neighborhood.

Second, the quotes could contain anything. For example "conjunctions eq 'and or xor not'". And because of that, you will need to track the quotes, so you don't split when you are within the quotes.

Vani Kumar wrote:
example 3:
"(address/zipcode eq '11122' or address/zipcode eq '22233') and (name/lastname eq 'xang' or name/firstname eq 'john') and status eq 'active' and contains(familyName, 'smith')";

And third, and this is a big one... you have to deal with precedence. In this example, the "or" within the parenthesis has higher precedence that the "and" outside of the parenthesis. You can't just split it directly. It has to be split into the parens components, and then when processing the parens components, it has to be split again (recursively).

Henry

[Vani Kumar]

Seems like split from string or commons stringutils (i tried to check if substringsBetween(string,"(",")") , subssubstringBefore work, but will miss the precedence operators like and, or) might not work for this case. Any other ideas, please suggest.

Right now i might not care about precedences of operators (i think may be the criteria in jpa can handle when i add all the expressions to it, not sure still, i'll double check)

[Campbell Ritchie]

I suspect that precedences and associativities would take your expressions out of the realms of regular grammars and into the realms of context‑free grammars, so you would need a proper parser as regular expressions can only parse regular grammars.

[Vani Kumar]

Don't have quite experience on parser and regular expressions as well. Please suggest any ideas. Thanks!

Just found some parser generators online, will do some research, learn and will check my luck.

[Campbell Ritchie]

You need to start by defining a grammar for your language.

[fred rosenberger]

Campbell Ritchie wrote:You need to start by defining a grammar for your language.

I would say you need to define the problem. Every time someone makes a suggestion, you come back with "actually, I need to do THIS" or "actually, I don't care about THAT". So until you can clearly define what you do and don't want to do, there's not much more advice anyone can give you.

[Vani Kumar]

I realized this feature can be achieved by using libraries like java framework/library like Odata4j, Apache olingo.