Find string which contains X but does not contain XY

[Edoardo Briganti]

Hello,
I am operating Stata/SE 15.1 on a Windows Server.
I have a string variable, called object_description, and I want to create another string variable, called product_id, which will be equal to "XYZ" if object_description contains the character "SHIP".
To complicate the problem, variable object_description also contains observations with words such as: "LIGHTSHIP", "MEMBERSHIP" and "SHIPPING". I don't want my new variable product_id to be equal to "XYZ" if object_description contains such expressions. Therefore, I do the following:

gen  product_id = "XYZ" if strpos(object_description, "SHIP") & ( !(strpos(object_description, "MEMBER")) | !(strpos(object_description, "LIGHT")) | !(strpos(object_description, "PING")) )

I also tried:

gen  product_id = "XYZ" if strpos(object_description, "SHIP") & ( !(strpos(object_description, "MEMBERSHIP")) | !(strpos(object_description, "LIGHTSHIP")) | !(strpos(object_description, "SHIPPING")) )

Neither work.
In fact, after I run the code and I browse the dataset, I find that whenever object_description contains, say, "MEMBERSHIP", my new variable product_id is equal to "XYZ", as if Stata were not reading what follows the &-statement.

I am very confused. Any suggestion would be much appreciated.

Cordially
Edoardo

[Clyde Schechter]

You got your boolean algebra mixed up. Think about what happens when object_description contains "MEMBERSHIP". First, strpos(object_description, "SHIP") is non-zero, that is, true.

Now let's look at the second conjunct. !strpos(object_description, "MEMBER") will be false, because MEMBER does appear. So far, so good. But then we have the disjunct !strpos(object_description, "LIGHT"). Well, LIGHT does not appear in object_description, so !strpos(object_description) is true. Therefore when ORed together with !strpos(object_description, "MEMBER") you get true, and, of course, it remains true when the other disjuncts get ORed in.

So the whole -if- condition comes down to true & true, which is, wait for it..., true.

What you want is:

gen product_id = "XYZ" if strpos(object_description, "SHIP") & !strpos(object_description, "MEMBER") & !strpos(object_description, "LIGHT") & !strpos(object_description, "PING")

If you would prefer to see fewer ! characters, you can apply the distributive law, which, done correctly gives

gen product_id = "XYZ" if strpos(object_description, "SHIP") & !(strpos(object_description, "MEMBER") | strpos(object_description, "LIGHT") | strpos(object_description, "PING"))

[Edoardo Briganti]

What a silly mistake! Time for a break from working. At least it was not a coding error.
Thanks Clyde for correcting me.

Best,
Edoardo