Replacing Text: Some Simple Examples
As I hinted earlier, groups have an important role in text replacement. To search-and-replace, you'll need to specify a replacement pattern string. Each character in this string is interpreted literally, with one exception: the dollar sign ($) has a special meaning. To use this syntax, the dollar sign is followed by a number:
This says, "Replace me with the nth group that was tagged (recognized) in the search string." (I'll explain tagging in a moment.) At this point, the natural question comes up, "What if I want to specify a literal dollar sign?" As you might guess by now, you'll use two dollar signs:
Let's look at a simple example. Bear in mind that the following items are involved:
- Regular expression. Specifies the character(s) you want the search to find.
- Replacement pattern string. Specifies how the searched string will be translated into new text.
- Target string. The string to be "operated on"—that is, the input string.
- Output string. The characters output by the C++ regular expression library.
Assume that we want to replace all double occurrences of a word with a single occurrence of that word. The following regular expression specifies the double-word pattern:
The expression [a-zA-Z]+ specifies a word. The range [a-zA-Z] accepts any lowercase or uppercase letter. The repeat operator (+) matches as many consecutive characters in this range as it can; that is, it matches as many letters in a row as it can. Matching stops when some non-alphabetic character is read, such as a space, or when the end of the target string is read.
The use of parentheses places [a-zA-Z]+ into a group, and something interesting happens: At runtime, a match group is tagged (remembered). The rest of the regular expression in our example says to match the first group again:
We need to make a critical distinction at this point: \1 means "Match group number 1 again." It doesn't produce the same result as this:
This would mean, "Match any word, match a space, and then match any word again." It would match input text such as the following:
But that's not what we want. Instead, we want a word to be matched and then remembered, and then we want the program to attempt to match that same word again. That's what tagging does.
Consider the following target string:
The cat cat bites the dog dog.
Assume that the following regular expression is applied:
Using functions and datatypes I'll demonstrate in the next section, the program tries to match a word, which is a pattern of text matching [a-zA-Z]+. The first word in the example sentence, The, matches the pattern. Because the expression is in parentheses, it forms a group. So, at runtime, that found text is tagged.
The next sub-expression means "Match the text that was tagged in the first group":
Notice the use of the backslash, the escape character: This character takes away special meaning from characters such as these: * + ( ) and so on. However, the use of the backslash in this example is an exception to the general rule. In this case, the backslash adds special meaning rather than taking it away. The digit 1 by itself would mean "Match the digit 1." But in this case it means "Match the tagged text." Therefore, in this example, the next thing the search function will do is attempt to match a space followed by another occurrence of the word The. In this case, it would fail.
But the function then matches the word cat, which is tagged (remembered). The match then succeeds, as the function immediately reads a second occurrence of cat.
What do we want to use to replace each group of repeated words? You can specify any text here. Look at the target string again:
The cat cat bites the dog dog.
You could replace each group of repeated words with the word ERROR, although that doesn't seem very useful:
The ERROR bites the ERROR.
Instead, let's replace each instance of repeated words with just one copy of the word. That's what the following replacement string does. The text matched at runtime was tagged, so it can be reused in the replacement pattern:
Given this replacement string, each repeated-word substring is replaced with just one copy of the contents of group 1 (which in this case is the only group). So after the search-and-replace, the final result is this:
The cat bites the dog.
For the sake of illustration, consider that each double-word substring could have been replaced with three copies of the contents of group 1 (that is, three copies of the word). In that case, the replacement text would be as follows:
With this result:
The cat cat cat bites the dog dog dog.
Replacement Using C++ Functions
Now let's look at search-and-replace carried out in C++ code. The first thing to note is that the backslash is used here in the search pattern; but when you code C++ strings, you must use two backslashes to indicate a single backslash.
The code requires several things:
- Regular expression to use for searching
- String with replacement pattern
- Target string
Before we proceed to the C++ code, some preliminaries:
#include <regex> #include <string> using std::regex; using std::string; using std::regex_replace;
Here are the string and regex declarations:
regex re("([a-zA-Z]+) \\1"); // Find double word. string replacement = "$1"; // Replace with one word. string target = "The cat cat bites the dog dog.";
You may notice a seeming inconsistency. The search pattern is declared as a special regex object, while the replacement pattern is just a standard string. That's just the way it works. The search pattern for any regex function is always a regex object. This design causes the program to build a special binary object, optimized for use at runtime. The replacement pattern is not built into a binary object until the function is called at runtime.
Finally, the regex_replace function carries out the search-and-replace. The action is to search for every sub-expression that matches the regex pattern (re) and then replace each occurrence with the replacement pattern:
string output_str = regex_replace(target, re, replacement);
If this string is displayed, the contents are now as follows:
The cat bites the dog.
Another Search-and-Replace Example
The previous examples were simplistic, but you should be able to see how complex text translations can be built upon these basic tools. In particular, you can tag and reuse any number of groups.
Consider a slightly more advanced usage. We could search for each substring having this form:
word1 bites word2
and replace it with this:
word2 bites word1
In other words, we're reversing the order. The following declarations set up this operation:
#include <regex> #include <string> using std::regex; using std::string; using std::regex_replace; regex re("([a-zA-Z]+) bites ([a-zA-Z]+)"); string replacement = "$2 bites $1";
Two groups are specified in the regular expression (re); therefore, the replacement string can legally refer to up to two groups. In this case, the order of the two words is switched, while the word "bites" is preserved as is. We could even have made "bites" a group, although nothing would be gained in this case. (It would involve some unnecessary extra operations at runtime, although the hit on efficiency would be small.) In that case, the same replacement could have been written this way:
regex re("([a-zA-Z]+) (bites) ([a-zA-Z]+)"); string replacement = "$3 $2 $1";
With these three groups, the first and third words are switched.
Either way, the code performing the search-and-replace operation looks something like this:
string replacement = "The cat bites dog, the dog bites flea." string output_str = regex_replace(target, re, replacement);
If the output string is then displayed, the result is the following:
The dog bites cat, the flea bites dog.