Extracting a string that occurs before an underscore

[Tan Nguyen]

Hi,

I have data as follow:

* Example generated by -dataex-. To install: ssc install dataex
clear
input str7 unique_id
"2044_18"
"205_16"
"207_8"
"2037_15"
"2037_14"
"2037_26"
"2044_25"
"2095_8"
"2095_2"
"2095_19"
"2044_28"
"2083_6"
"2037_16"
"2037_9"
"2037_23"
"2037_19"
"2074_8"
"2037_10"
"2037_21"
"2037_11"
"2044_4"
"2083_13"
"2041_11"
"2041_24"
"2044_21"
"2083_10"
"2090_23"
end

I want to create a new variable to take value before an underscore like 2044, 205, 207.

Thanks,
Tan

[Rich Goldstein]

try the following 2 commands:

gen byte pos=strpos(unique_id, "_")
gen newvar=substr(unique_id,1, pos-1)

you can then keep or drop "pos" and you see fit; you can choose whatever name you want instead of the "newvar" that I used - and, of course, you can -destring- it if you want

[Hemanshu Kumar]

Assuming that the part before the underscore is always digits, here is some alternative code using regular expressions:

gen int wanted = real(ustrregexs(1)) if ustrregexm(unique_id,"(\d+)_")

[Bjarte Aagnes]

ref #5 ignore the following error

gen long firstpart = real(subinstr(unique_id), "_", ".", 1)

correction:

gen long firstpart = real(subinstr(unique_id, "_", ".", 1))

gen firstpart = int(real(subinstr(unique_id, "_", ".", 1)))

[Hemanshu Kumar]

Bjarte Aagnes can you help me understand how #4 manages to work?

One part that seems crucial is setting the data type to be long, so that the non-integer part is dropped.

What I'm confused about is, the line as you entered seems to suggest that subinstr() is a function that takes one argument, while real() takes four, while it is the opposite. How is that code even legal, instead of needing to be

gen long firstpart = real(subinstr(unique_id, "_", ".", 1))

[Hemanshu Kumar]

Bjarte Aagnes thanks for the edits to #4, but I'm still confused -- your original code actually works, whereas I would have thought it's syntactically wrong. I have no clue how it manages to work!

Again, the original code is:

gen long firstpart = real(subinstr(unique_id), "_", ".", 1)

[Bjarte Aagnes]

I report this issue to tech support:

display real(subinstr("2044_18"), "_", ".",1) // should fail like the mata code:
mata : strtoreal(subinstr("2044_18"), "_", ".", 1)

[Hemanshu Kumar]

Funnily enough:

. display real(subinstr("2044_18"), "_", ".",1)
2044.18

but

. display real(subinstr("2044_18"), "_", ".",1) // should fail like the mata code:
unknown function ()
r(133);

[Bjarte Aagnes]

#8 is probably the result of running the command in the command window where the /* */, //, and /// comment indicators cannot be used. The error message should be more precise and informative, and consistent, if illegal commenting is used interactively.

[Hemanshu Kumar]

You're absolutely right. From a do-file, the result of both commands in #8 is identical, 2044.18. My bad.