Extract phone numbers from text #252

AndyGauge · 2017-07-19T21:46:21Z

Fixes #241

budziq

Hi @AndyGauge nice work! unfortunately I'll not be able to give you a proper review as Im mobile only on holidays so here are some quick thoughts.

Please update the PR description to make the related issue autoclose on merge ("fixes" clause)
If the example will read an external file please make it no_run

budziq · 2017-07-19T22:15:36Z

src/basics.md

+    println!("Input a phone number");
+    let mut phone_input = String::new();
+    // Regular expression is looking for 3 groups of digits, capturing them
+    let re = Regex::new(r".*([2-9]\d{2}).*(\d{3}).*(\d{4})")?;


hi this regex will catch false positives "fiz 233 baz 345 bar 6789" so this might not be the best one.
on the other hand the allowed NANP spec is ridiculously complex and even a subset of it spawns not very nice:
^(\+\d{1,2}\s)?$?\d{3}$?[\s.-]?\d{3}[\s.-]?\d{4}$ (copied via mobile so it might vie utterly broken)

So lets meet halfway and agree that phone numbers are optional country region code + 3 groups of digits optionally separated by either single space dot or -. Country region code is either preceeded by + or 00 If its still too complex we might pair it down further as long as it will not give false positives.

Also it might be a good place to showoff named capture groups (at least for the country code)

And we would still mention in the example that this will work only for nicely formatted numbers.

So I've investigated international phone numbers and their format is way different. Here's the wikipedia article I'm investigating: https://en.wikipedia.org/wiki/National_conventions_for_writing_telephone_numbers
Because it would be easy to create an entire library out of the individual nation's formatting requirements I want to make this only function for US telephone numbers. (While including the optional +1). International conventions includes dots, dashes, spaces, and parenthesis optionally between digits. I think +15551231234 and 5551231234 are both valid phone numbers. Just that regex alone is actually a bit more complicated than the one you posted via mobile. (and won't use the ^and$ because we are not validating we are extracting).

Anyway I'll throw something together that eliminates the false positives while looking good enough to reason (while only guaranteeing support for US).

Ok fair enough 👍 just leave a comment that it'll match only US convention.

budziq · 2017-07-19T22:23:28Z

src/basics.md

+    let mut phone_input = String::new();
+    // Regular expression is looking for 3 groups of digits, capturing them
+    let re = Regex::new(r".*([2-9]\d{2}).*(\d{3}).*(\d{4})")?;
+    stdin().read_line(&mut phone_input)?;


I would suggest to open some arbitrary text file and iterate over all of its lines and phone numbers.

Also we might want to collect the numbers into a set in order to deduplicate them prior to printing.

Good ideas, but should we just use a static string so that it works in a cookbook?

you can either provide a short static string (then assert_eq on the results to show the intended output) or a filename (in that case make the example no_run)

budziq · 2017-07-19T22:24:33Z

src/basics.md

+        let formatted_phone = format!("({}) {}-{}", &captures[1], &captures[2], &captures[3]);
+        println!("Formatted like a boss: {}", formatted_phone);
+    } else {
+        println!("Phone numbers must include area code.  Use numbers and whatever else you want\nFor example +1 (555) 101-9939");


lets remove the lines that are not really to the point

budziq · 2017-07-19T22:26:10Z

src/basics.md

+    // Make sure the regular expression captured something
+    if let Some(captures) = re.captures(&phone_input){
+        let formatted_phone = format!("({}) {}-{}", &captures[1], &captures[2], &captures[3]);
+        println!("Formatted like a boss: {}", formatted_phone);


how about just printing these out instead of formatting? also no need to add additional noise to the output. lets keep the example short and sweet so it can be copy pasted.

Also I leave it up to you if you would consider using PhoneNumber struct with display (please see the external process example)

AndyGauge · 2017-07-21T17:45:38Z

src/basics.md

+            number: [cap[1].to_string(),cap[2].to_string(),cap[3].to_string()].concat()
+        }
+    });
+    assert!(format!("{}", phone_numbers.next().unwrap()) == "+1 (505) 881-9292");


Is there any way other than unwrap for this option?

yep assert_eq!(phone_numbers.next(), Some("+1 (505) 881-9292"))
possibly you will need to do some to_str or somesuch

When I do that I get this:

error[E0308]: mismatched types --> src/main.rs:57:5 | 57 | assert_eq!(phone_numbers.next(), Some("+1 (299) 339-1020")); | ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ expected struct `PhoneNumber`, found &str | = note: expected type `std::option::Option<PhoneNumber>` found type `std::option::Option<&str>` = help: here are some functions which might fulfill your needs: - .take() - .unwrap() - .unwrap_or_default() = note: this error originates in a macro outside of the current crate

I could do this

assert_eq!(phone_numbers.next(), Some(PhoneNumber {country:"US".to_string(), number:"2993391020".to_string()}));

but that doesn't look any better. I really think if were are trying to offer something someone can use, implementing Display and showing how to have the phone numbers formatted is what people can use.

that is why I've suggested something like
assert_eq!(phone_numbers.next().map(String::from), Some("+1 (299) 339-1020".to_owned()));

budziq

@AndyGauge almost there. just few minor changes and git squash to a single commit and we good to go!

budziq · 2017-07-21T18:21:03Z

src/basics.md

+                         (\d{4})                    #Subscriber Number")?;
+    let mut phone_numbers = re.captures_iter(phone_text).map(|cap| {
+        PhoneNumber {
+            country: "US".to_owned(),


the country does not make much sense here as its unconditionally "US" maybe lets just drop it (also using String as a match tag would not be a best idea).

how about storing area exchange and subscriber codes separately instead?

budziq · 2017-07-21T18:22:27Z

src/basics.md

+    fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
+        match self.country.as_ref() {
+            "US" => write!(f, "+1 ({}) {}-{}", &self.number[0..3], &self.number[3..6], &self.number[6..10]),
+            _    => write!(f, "{}", self.number),


lets just drop the country and ommit the ±1 from printing . the regex should still conditionally match it

budziq · 2017-07-21T18:24:24Z

src/basics.md

+#
+struct PhoneNumber {
+    country: String,
+    number: String,


maybe the PhoneNumber can refer to not owned str?

budziq · 2017-07-21T18:27:27Z

src/basics.md

+    1.299.339.1020";
+    let re = Regex::new(r"(?x)
+                         \+?1?                      #Country Code Optional
+                         [\x20\(\.]*


the arbitrary number of parentheses matchers look funky.
we probably shoud match areacode in two alternatives one with one set of paren and other with none.

also there are better ways to match whitespace than 20 ascii code.

\x20 is required with the (?x) ignore whitespace. I could use \s but that would match tabs and newlines. If you are OK with that I can switch it.

Is this what you mean?

[\s\.]*|(?\s\()?

\s?[\.\(]?

yeah I would go with \s

\+?1? should be moved int an optional group (\+1)?\s* so that plus char has to be followed by 1otherwise these two are independent

I meant that the area codes expression within the parenthesis has to be matched with both opening and closing parens as a whole or | without both of them. currently the match for opening ,closing parens are independent so we can match (((202 991 9534 .

I may try to help but writing regexes on mobile is a real pain and I have no way to give you the full answer or even verify if my suggestions are fully sane. So you might try to solving this yourself or wait for me to get back behind a keyboard within next two weeks 😢 .

budziq · 2017-07-21T18:29:59Z

src/basics.md

+}
+#
+# fn main() {
+#    run();


please use quick_main!

budziq · 2017-07-21T18:36:24Z

src/basics.md

+## Extract phone numbers from text
+
+Attempts to process a static string as a list including US phone numbers.  The
+expression is checking that there are 7 digits that come in groups of 3 3 4,


I would not try to explain the actual regex here. we are interested in the crate and its methods and don't try to teach regular expressions

budziq · 2017-07-21T18:37:51Z

src/basics.md

+Attempts to process a static string as a list including US phone numbers.  The
+expression is checking that there are 7 digits that come in groups of 3 3 4,
+separated by spaces, dots, dashes, and parenthesis.  The phone number is then
+tested based on formatting rules.


the fact that we are testing is self evident from the asserts in the code.

budziq · 2017-07-21T18:42:11Z

src/basics.md

+<a name="ex-phone"></a>
+## Extract phone numbers from text
+
+Attempts to process a static string as a list including US phone numbers.  The


This sentence looks a little off, I wouldn't say that the code "attempts" it actually does the parsing. Also I would keep the whole description as a short one sentence mentioning the key method used (captures_iter).

budziq · 2017-07-21T18:56:02Z

hi I no longer see your comment on github. I'm currently on mobile for the next 1.5weeks so I'll not be able to help with the finer details. In general you would map the result on the left with to_string and on the right you would leave Some("the number str".to_owned()) no need to call fmt explicitly or unwrap. the assert_eq is able to compare complex structs such as Option<String> or something along those lines. there might be a nicer way but away from the keyboard I cant give a better suggestion from the top of my head. ay a little with rustc and you'll get it right 👍

…

On 21 Jul 2017 20:43, "Andrew Gauger" ***@***.***> wrote: ***@***.**** commented on this pull request. ------------------------------ In src/basics.md <https://github.com/brson/rust-cookbook/pull/252#discussion_r128833949>: > + 1.299.339.1020"; + let re = Regex::new(r"(?x) + \+?1? #Country Code Optional + [\x20$\.]* + ([2-9]\d{2}) #Area Code + [\x20$\.\-]* + ([2-9]\d{2}) #Exchange Code + [\x20\.\-]* + (\d{4}) #Subscriber Number")?; + let mut phone_numbers = re.captures_iter(phone_text).map(|cap| { + PhoneNumber { + country: "US".to_owned(), + number: [cap[1].to_string(),cap[2].to_string(),cap[3].to_string()].concat() + } + }); + assert!(format!("{}", phone_numbers.next().unwrap()) == "+1 (505) 881-9292"); So when I do that I get error[E0308]: mismatched types --> src/main.rs:53:5 | 53 | assert_eq!(phone_numbers.next().unwrap().fmt(), "+1 (505) 881-9292 <(505)%20881-9292>"); | ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ expected enum `std::result::Result`, found &str | = note: expected type `std::result::Result<(), std::fmt::Error>` found type `&str` = note: this error originates in a macro outside of the current crate I'm using the Some value and then running it through format! to show the output of the struct. Maybe I need to re-write? I thought the concern about unwrap is that it may panic, but that's the point of an assert--to panic when the values aren't equal. — You are receiving this because you commented. Reply to this email directly, view it on GitHub <https://github.com/brson/rust-cookbook/pull/252#discussion_r128833949>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AANfSGyUv2bmCjSFQudnaCHmk2-e2-gAks5sQPFGgaJpZM4OdVwh> .

budziq · 2017-07-21T22:18:20Z

src/basics.md

+
+fn run() -> Result<()> {
+    let phone_text = "
+    +1 505 881 9292


also this code should allow for multiple numbers per line so let's show it

budziq

looking better!

budziq · 2017-07-21T23:31:17Z

src/basics.md

+            number: [area_code,cap[4].to_string(),cap[5].to_string()].concat()
+        }
+    });
+    assert_eq!(phone_numbers.next().map(|m| format!("{}", m)), Some("1 (505) 881-9292".to_owned()));


you can just .map(|m| m.to_string() this is a preferred variant (clippy has a lint for it)

budziq · 2017-07-21T23:35:34Z

src/basics.md

+#
+#[derive (PartialEq, PartialOrd, Debug)]
+ enum CountryCode {
+     US,


please see my previous comment about country code not doing any real work. it is mostly noise obfuscating a quite simple and elegant example. i would just remove it (the part about matching on Strings was just a sidenote ;) sorry if i was unclear)

budziq · 2017-07-21T23:37:48Z

src/basics.md

+
+#[derive (PartialEq, PartialOrd, Debug)]
+struct PhoneNumber {
+    country: CountryCode,


please remove the country code but we might like to store the area and exchange codes separately (my comments from previous patcset)

budziq · 2017-07-21T23:39:02Z

src/basics.md

+struct PhoneNumber {
+    country: CountryCode,
+    number: String,
+}


please see my comments about making this zero-copy and storing &str instead String

I have tried to do this, but I cannot figure out the lifetime. Here's the problem:

error[E0597]: borrowed value does not live long enough --> src\main.rs:49:5 | 47 | number: &[area_code,cap[4].to_string(),cap[5].to_string()].concat() | ---------------------------------------------------------- temporary value created here 48 | } 49 | }); | ^ temporary value dropped here while still borrowed ... 57 | } | - temporary value needs to live until here

Here's the struct def:

#[derive (PartialEq, PartialOrd, Debug)] struct PhoneNumber<'a> { number: &'a str, } // Allows printing phone numbers based on country convention. impl<'a> fmt::Display for PhoneNumber<'a> { fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result { write!(f, "1 ({}) {}-{}", &self.number[0..3], &self.number[3..6], &self.number[6..10]) } }

Can you advise?

There is no clean way of concatenating the str's without allocation

The easiest way would be to follow my suggestion of separate &str member for each capture (area, extension, subscriber). Then you don't need to concat and allocate

budziq · 2017-07-21T23:39:36Z

src/basics.md

+impl fmt::Display for PhoneNumber {
+    fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
+        match self.country {
+            CountryCode::US => write!(f, "1 ({}) {}-{}", &self.number[0..3], &self.number[3..6], &self.number[6..10]),


please run rustfmt on the snippet

budziq · 2017-07-22T09:36:31Z

src/basics.md

+    1 (800) 233-2010
+    1.299.339.1020";
+    let re = Regex::new(r"(?x)
+                        (?\+?1)?                         # Country Code Optional


the first question mark within the parens looks off but I cannot verify it right now so I might be wrong. the rest looks really nice!

AndyGauge · 2017-07-22T22:20:35Z

src/basics.md

        PhoneNumber {
-            country: CountryCode::US,
-            number: [area_code,cap[4].to_string(),cap[5].to_string()].concat()
+            area: if cap.get(2) == None { &cap[3] } else { &cap[2] },


error[E0597]: `cap` does not live long enough --> src\main.rs:56:44 | 56 | area: if cap.get(2) == None { &cap[3] } else { &cap[2] }, | ^^^ does not live long enough ... 60 | }); | - borrowed value only lives until here | = note: borrowed value must be valid for the static lifetime...

Same problem. cap only exists inside the iter() block,

probably we want to use as_str instead of depending on &capture Deref into a &str. Pease see a soon to be merged PR for suggestions https://github.com/brson/rust-cookbook/pull/247/files

also I would suggest using match on captures instead of imperative testing for None

Anyhow if its still a blocker then just go with owned version for now and we'll update the example in a week when I'm back behind a keyboard 👍

brson · 2017-07-25T00:55:46Z

I rebased and merged. Thanks @AndyGauge

budziq requested changes Jul 19, 2017

View reviewed changes

AndyGauge commented Jul 21, 2017

View reviewed changes

budziq requested changes Jul 21, 2017

View reviewed changes

budziq reviewed Jul 21, 2017

View reviewed changes

AndyGauge force-pushed the ex-phone branch 3 times, most recently from 2f30cd0 to 13483b3 Compare July 21, 2017 23:10

budziq requested changes Jul 22, 2017

View reviewed changes

AndyGauge commented Jul 22, 2017

View reviewed changes

AndyGauge force-pushed the ex-phone branch from 1d8d68c to 9afc8e9 Compare July 23, 2017 01:04

Extract phone numbers from text

9afc8e9

brson closed this Jul 25, 2017

Extract phone numbers from text #252

Extract phone numbers from text #252

Conversation

AndyGauge commented Jul 19, 2017 • edited

budziq left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

budziq Jul 21, 2017 • edited

Choose a reason for hiding this comment

Choose a reason for hiding this comment

budziq Jul 21, 2017 • edited

Choose a reason for hiding this comment

budziq left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

AndyGauge Jul 21, 2017 • edited

Choose a reason for hiding this comment

AndyGauge Jul 21, 2017 • edited

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

budziq commented Jul 21, 2017 via email

Choose a reason for hiding this comment

budziq left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

AndyGauge Jul 22, 2017 • edited

Choose a reason for hiding this comment

budziq Jul 22, 2017 • edited

Choose a reason for hiding this comment

brson commented Jul 25, 2017

AndyGauge commented Jul 19, 2017 •

edited

budziq Jul 21, 2017 •

edited

budziq Jul 21, 2017 •

edited

AndyGauge Jul 21, 2017 •

edited

AndyGauge Jul 21, 2017 •

edited

AndyGauge Jul 22, 2017 •

edited

budziq Jul 22, 2017 •

edited