Simple String Patterns

This library makes it easier to match, split and extract strings in Rust. It builds on the Rust standard library. A parallel string-patterns crate provides extensions to work with regular expressions. Together, these crates aim to make working with strings as easy in Rust as it is Javascript or Python with cleaner syntax. Simpler string matching methods such as starts_with, contains or ends_with will always perform better, especially when processing large data sets. Methods such as equals_ci, equals_ci_alphanum, starts_with_ci and starts_with_ci_alphanum build on these core methods to facilitate string manipulation without regular expressions.

In version 0.4.0, only the core methods are included by default. String bounds rule sets and the fluent match chain are both opt-in features, and other supplementary functions have been transferred to other library crates.

A compelling case for this crate is small microservices and CLI utilities that parse predictable, pre-formatted text output from other tools or services — log lines, status messages, command output — where a regex is overkill and manually normalizing case and punctuation by hand is repetitive boilerplate. It's tempting to reach for contains() to spot a severity level in a log line, but that gives false positives whenever the word shows up incidentally in the message body rather than as the actual tag:

let warning_msg = "[warning] high CPU usage. This may cause errors.";
let error_msg = "[error] cannot divide by 0. You have been warned";

warning_msg.contains_ci("error"); // true — WRONG, this is a warning, not an error
error_msg.contains_ci("warn");    // true — WRONG, this is an error, not a warning

Anchoring the check with starts_with_ci_alphanum instead of scanning the whole message avoids that, since it only looks at the tag, not the prose that follows:

warning_msg.starts_with_ci_alphanum("warning"); // true
warning_msg.starts_with_ci_alphanum("error");   // false
error_msg.starts_with_ci_alphanum("error");     // true
error_msg.starts_with_ci_alphanum("warning");   // false

It's also more robust than an exact literal prefix check, since it strips punctuation from both sides before comparing — so it survives the tag's delimiters changing (e.g. a log format moving from [warning] to **warning**:) without needing to be updated:

let reformatted = "**warning**: High CPU usage";
reformatted.starts_with_ci("[warning]");        // false — breaks: the delimiters changed
reformatted.starts_with_ci_alphanum("warning"); // true — still works

which in turn is more direct than doing the equivalent normalization by hand:

if input_string
    .to_lowercase()
    .chars()
    .filter(|c| c.is_alphanumeric())
    .collect::<String>()
    .starts_with("warning") {
  // ...
}

and, as of the 0.4.0 ci_engine rewrite, no longer pays for building that intermediate String on every call either — see Simple Patterns versus Regular Expressions below. Above all, it simply avoids the need to add regex as a dependency at all for a lightweight app that only has to check log files or call out to other CLI tools — no extra compile time or binary size for a job this small.

Features

The crate core is always just the SimpleMatch, MatchOccurrences and SimplContainsType traits (in matches.rs) — no other functionality compiles without opting in via Cargo features. Neither rules nor chain is enabled by default.

simple-string-patterns = { version = "0.4", features = ["rules", "chain"] }

Dependencies

• alphanumeric — Provides IsNumeric, StripCharacters, and CharGroupMatch traits for international number format parsing, character filtering, and numeric string extraction

Supplementary crates

The string-splitting and enclosing functionality previously bundled in this crate now lives in standalone crates. They are not dependencies of simple-string-patterns, but pair well with it if you need that functionality too:

• to_segments — The ToSegments trait for ergonomic string splitting with readable methods for common manipulation tasks.
• enclose-strings — Wrap or enclose strings in matching or complementary characters with optional escaping.

Simple Patterns versus Regular Expressions

simple-string-patterns aims to improve readability and minimise overhead in lightweight applications that would not otherwise need regex support. Under the hood, regular expression engines compile regex syntax and convert them into more efficient string matching subroutines.

Benchmarked against a compiled regex over 200,000 synthetic log lines ("line".starts_with_ci_alphanum("float") && "line".contains_ci("error") vs /^[^a-z0-9]*float.*error/i): the two are now close, ~29ns/line for this crate's methods versus ~15ns/line for the compiled regex — a modest, not dramatic, gap. starts_with_ci/ends_with_ci/equals_ci can bail out after inspecting only a handful of characters and are the strongest case for this crate; contains_ci alone has to scan the whole string and is the part of that gap that remains, since regex's substring search (two-way algorithm plus a memchr-based literal prefilter) is hard to beat without reimplementing it. If you need to add multiple nested rules or complex patterns, a regex is likely to stay faster and more flexible — the sibling regex-powered string-patterns crate makes that easy, and as of its own 0.4.0 shares a single compiled-once regex cache across all its matching methods, so "compile once, apply to many rows" is no longer something you have to build yourself in either crate. This crate is best suited to small utilities that need to process large quantities of strings with a range of highly predictable formats, e.g. in cryptography, logging.

Method overview

Simple case-insensitive match

let str_1 = "Dog food";
if str_1.starts_with_ci("dog") {
  println!("{} is dog-related", str_1);
}

Simple case-insensitive match on the alphanumeric characters only in a longer text

// Theses methods are handy for validating text values from external data sources with
// inconsistent naming conventions, e.g. first-name, first_name, firstName or "first name"
let str_1 = "Do you spell hip-hop with a hyphen?";
if str_1.contains_ci_alphanum("hiphop") {
  println!("{} is hip-hop-related", str_1);
}
// Compare a single term with or without hyphens, spaces or other punctuation
let sample_str = "Start-up";
if str_1.equals_ci_alphanum("startup") {
  println!("It's a start-up company");
}

Filter a vector of strings by their first alphanumeric characters

// Methods ending in _alphanum are good for filtering strings that may have other
// punctuation or casing inconsistencies around the part you care about
let sample_strs = [
  "/blue-sky.jpg",
  "----bluesky.png",
  "-B-l-u-e--sky",
  "Blueberry",
  " Blue sky thinking"
];
let strings_starting_with_blue = sample_strs
  .into_iter()
  .filter(|s| s.starts_with_ci_alphanum("bluesky"))
  .collect::<Vec<&str>>();
// should return all except "Blueberry"

Chain several checks on one string (chain feature)

// Requires the optional "chain" feature. Unlike contains_ci() and starts_with_ci_alphanum(),
// which return a plain bool and so cannot be chained further, match_contains_ci() returns a
// StringMatch that carries the accumulated result and the subject string through the chain.
let sample_1 = "Floating-point error at line 667";
let sample_2 = "[floating-point] error in line 999";
let sample_3 = "Rounding error at line 12";

let is_float_error = |s: &str| s.match_contains_ci("error")
  .and_starts_with_ci_alphanum("float")
  .result();

assert!(is_float_error(sample_1));
assert!(is_float_error(sample_2));
assert!(!is_float_error(sample_3));

// and_not_* and or_* combinators are also available, e.g. to exclude a suffix:
let is_cat_but_not_psd = "cat-picture.jpg".match_starts_with_ci("cat")
  .and_not_ends_with_ci(".psd")
  .result(); // true

Extract the first decimal value as an f64 from a longer string

const GBP_TO_EURO: f64 = 0.835;
let sample_str = "Price £12.50 each";
if let Some(price_gbp) = sample_str.to_first_number::<f64>() {
    let price_eur = price_gbp / GBP_TO_EURO;
    println!("The price in euros is {:.2}", price_eur);
}

Extract numeric sequences from phrases and convert them to a vector of floats

// This crate uses the alphanumeric dependency for advanced number extraction
// See https://crates.io/crates/alphanumeric for international number format support
let sample_str = "2.500 grammi di farina costa 9,90€ al supermercato.";
let numbers: Vec<f32> = sample_str.to_numbers();
// If two valid numbers are matched assume the first is the weight
if numbers.len() > 1 {
  let weight_grams = numbers[0];
  let price_euros = numbers[1];
  let price_per_kg = price_euros / (weight_grams / 1000f32);
  println!("Flour costs €{:.2} per kilo", price_per_kg);
}

Split a string list of numbers into floats

// extract 64-bit floats from a comma-separated list
// numbers within each segment are evaluated separately
let sample_str = "34.2929,-93.701";
let numbers = sample_str.split_to_numbers::<f64>(",");
// should yield vec![34.2929,-93.701]; (Vec<f64>)

Match by all or any pattern rules without regular expressions (rules feature)

// Call .as_vec() to convert the BoundsBuilder to a slice of StringBounds rules
let mixed_conditions = bounds_builder()
  .containing_ci("nepal")
  .ending_with_ci(".jpg");
let sample_name_1 = "picture-Nepal-1978.jpg";
let sample_name_2 = "edited_picture-Nepal-1978.psd";
// contains `nepal` and ends with .jpg
sample_name_1.match_all_conditional(&mixed_conditions.as_vec()); // true
// contains `nepal` but does not end with .jpg
sample_name_2.match_all_conditional(&mixed_conditions.as_vec()); // false
// contains `nepal` and/or .jpg
sample_name_1.match_any_conditional(&mixed_conditions.as_vec()); // true
// contains `nepal` and/or .jpg
sample_name_2.match_any_conditional(&mixed_conditions.as_vec()); // true

Filter by all pattern rules without regular expressions (rules feature)

// The same array may also be expressed via the new bounds_builder() function with chainable rules:
// You may call .as_vec() to convert to a vector of StringBounds rules as used by methods ending in _conditional
let mixed_conditions = bounds_builder()
  .containing_ci("nepal")
  .not_ending_with_ci(".psd");
let file_names = [
  "edited-img-Nepal-Feb-2003.psd",
  "image-Thailand-Mar-2003.jpg",
  "photo_Nepal_Jan-2005.jpg",
  "image-India-Mar-2003.jpg",
  "pic_nepal_Dec-2004.png"
];
/// The filter_all_rules() method accepts a *BoundsBuilder* object.
let nepal_source_files: Vec<&str> = file_names.filter_all_rules(&mixed_conditions);
// should yield two file names: ["photo_Nepal_Jan-2005.jpg", "pic_nepal_Dec-2004.png"]
// This will now return Vec<&str> or Vec<String> depending on the source string type.

Nested Rule Sets (rules feature)

As of version 0.3.0 you may add nested rule sets with and / or logic. The former case is true only if all conditions are met, while the latter is true if any of the conditions are met. The BoundsBuilder struct now has a set of methods starting with and or or. You may call and(rules: BoundsBuilder) or or(rules: BoundsBuilder) directly with a nested rule set if you have a mix of rule types. However, if all rules have the same bounds, other methods accepting a simple array of patterns are available, e.g.

• or_starting_with_ci(patterns: &[&str])
• or_starting_with_ci_alphanum(patterns: &[&str])
• or_containing_ci(patterns: &[&str])
• or_ending_with_ci(patterns: &[&str])
• and_not_ending_with_ci(patterns: &[&str])

let filenames = [
  "my_rabbit_2019.webp",
  "my_CaT_2020.jpg",
  "neighbours_Dog_2021.gif",
  "daughters_Dog_2023.jpeg",
  "big cat.psd"
];
/// Match files containing the letter sequences "cat" or "dog" and ending in ".jpg" or ".jpeg";
let rules = bounds_builder()
  .or_containing_ci(&["cat", "dog"])
  .or_ending_with_ci(&[".jpg", ".jpeg"]);
let matched_files = filenames.filter_all_rules(&rules);
/// Should yield an array with "my_CaT_2020.jpg" and "daughters_Dog_2023.jpeg"

The above example reproduces the following example regular expression /(cat|dog).*?.jpe?g$/. The _alphanum-suffixed variants let match only on numbers and letters within a string, i.e. ignorning any spaces or punctuation.

Filter by any pattern rules without regular expressions (rules feature)

// The same array may also be expressed via the new bounds_builder() function with chainable rules:
// Call .as_vec() at the end
let mixed_or_conditions = bounds_builder()
  .containing_ci("nepal")
  .containing_ci("india");
let file_names = &[
  "edited-img-Nepal-Feb-2003.psd",
  "image-Thailand-Mar-2003.jpg",
  "photo_Nepal_Jan-2005.jpg",
  "image-India-Mar-2003.jpg",
  "pic_nepal_Dec-2004.png"
];
let nepal_and_india_source_files: Vec<&str> = file_names.filter_any_rules(&mixed_or_conditions);
// should yield two file names: ["edited-img-Nepal-Feb-2003.psd", "photo_Nepal_Jan-2005.jpg", "image-India-Mar-2003.jpg", "pic_nepal_Dec-2004.png"]
// To combine and/or logic, you can filter all rules with a nested "or" clause.
let mixed_conditions_jpeg_only = bounds_builder()
  .ending_with_ci(".jpg")
  .or(mixed_or_conditions);
let nepal_and_india_source_files_jpgs: Vec<&str> = file_names.filter_all_rules(&mixed_conditions_jpeg_only);
// should yield two file names: ["photo_Nepal_Jan-2005.jpg", "image-India-Mar-2003.jpg"]

Filter strings by character categories

let sample_str = "Products: $9.99 per unit, £19.50 each, €15 only. Zürich café cañon";
let vowels_only = sample_str.filter_by_type(CharType::Chars(&['a','e','i','o', 'u', 'é', 'ü', 'y']));
println!("{}", vowels_only);
// should print "oueuieaoyüiaéao"
let lower_case_letters_a_to_m_only = sample_str.filter_by_type(CharType::Range('a'..'n'));
println!("{}", lower_case_letters_a_to_m_only);
// should print  "dceieachlichcafca"
/// You can filter strings by multiple character categories
let sample_with_lower_case_chars_and_spaces = sample_str.filter_by_types(&[CharType::Lower, CharType::Spaces]);
println!("{}", sample_with_lower_case_chars_and_spaces);
// Should print "roducts  per unit  each  only ürich café cañon"

Strip spaces only

let sample_str = "19 May 2021 ";
let sample_without_spaces = sample_str.strip_spaces();
println!("{}", sample_without_spaces);
  // should print "19May2021";

Remove character categories from strings

let sample_without_punctuation = sample_str.strip_by_type(CharType::Punctuation);
println!("{}", sample_without_punctuation);
// should print "Products 999 per unit £1950 each €15 only Zürich café cañon";
let sample_without_spaces_and_punct = sample_str.strip_by_types(&[CharType::Spaces, CharType::Punctuation]);
println!("{}", sample_without_spaces_and_punct);
// should print "Products999perunit£1950each€15onlyZürichcafécañon";

Traits

Enums

CaseMatchMode

Defines case-sensitivity and alphanumeric-only modes.

StringBounds

Defines simple match rules with the pattern and a positivty flag, e.g. StringBounds::Contains("report", true, CaseMatchMode::Insensitive) or StringBounds::EndsWith(".docx", CaseMatchMode::Insensitive). The bounds_builder method helps build these rule sets. All options have pattern: &str, is_positive: bool and case match mode flags and acceot the same three arguments (&str, bool, CaseMatchMode) for the match pattern, positivity and case match mode.

CharType

Defines categories, sets or ranges of characters as well as single characters.

Structs

BoundsBuilder

This struct helps you build string pattern rules for use with the matched_by_rules(), filter_all_rules() and filter_any_rules() methods. The bounds_builder() function returns a base instance on which you may chain any number of rules and sub-rules.

Dev Notes

This crate serves as a building block for other crates as well as to supplement a future version of string-patterns. Some updates reflect minor editorial changes.

Version 0.4.0 streamlines the crate, fixes a performance regression, and corrects two bugs

Crate structure. The ToSegments, ToSegmentFromChars, SimpleEnclose and ToStrings traits described in earlier versions of these notes have been removed from this crate and now live in the standalone to_segments and enclose-strings crates (see Supplementary crates above). The string bounds rule set (bounds_builder(), StringBounds, and the _rules/_conditional methods) is now gated behind the rules feature, and a new chain feature adds StringMatch, a fluent match_*/and_*/or_* chain for combining several checks on one string (see Features and the chaining example above) — neither feature is enabled by default. By default the crate now exposes only SimpleMatch, MatchOccurrences and SimplContainsType.

Upgrading from 0.3.x: BoundsBuilder/bounds_builder()/StringBounds were always compiled in on 0.3.x. If you use them, add features = ["rules"] explicitly — without it, code that called bounds_builder() will fail to compile after upgrading to 0.4.0, not just silently lose behaviour.

Performance. The _ci/_ci_alphanum methods on SimpleMatch (contains_ci, starts_with_ci_alphanum, etc.) are now backed by an internal ci_engine module instead of the previous self.to_lowercase().strip_non_alphanum()... approach, which allocated a full transformed copy of the subject string on every call. The new implementation dispatches to a byte-level ASCII fast path (no allocation, no UTF-8 decoding) when both the subject and pattern are ASCII, falling back to a Unicode-correct character stream otherwise (café/naïve-style input is still handled correctly — see the tests in ci_engine.rs). Method names and bool return types are unchanged; only the internals got faster. See Simple Patterns versus Regular Expressions above for current numbers against a compiled regex, and benches/matching.rs (run with cargo bench) to reproduce them.

Bug fixes. starts_with_ci_alphanum, ends_with_ci_alphanum and contains_ci_alphanum now strip non-alphanumeric characters from both the subject and the pattern, matching equals_ci_alphanum's existing behaviour — previously only the subject was stripped, so a pattern containing punctuation or spaces behaved inconsistently across these methods. Separately, BoundsBuilder::and_starting_with_ci (rules feature) was silently installing the negated rule instead of the positive one it's named for, due to a copy-paste mistake; it now behaves as documented, and a regression test (test_and_starting_with_ci_positive) covers it.

Version 0.3.13 introduces the .strip_spaces() method as shorthand for .strip_by_type(CharType::Spaces). Version 0.3.11 introduces a .split_to_numbers::<T>(pattern: &str) method to split a string list of numbers into a vector of the specified number type. This is handy when parsing common input formats such as latitudes and longitudes represented as "42.282,-89.3938". This might fail via .to_numbers() when commas or points used as separators may be confused with decimal or thousand separators without other characters in between.

Version 0.3.8 New *andnot+ rules methods

This version introduced a set of andnot-prefixed rule methods to filter strings do not match the specified array of patterns, e.g. if we have a list image file names that start with animal names and we want to match those beginning with case-insensitive "cat" or "dog", but excluding those ending in "".psd" or ".pdf".

  /// file names starting with cat or dog, but not ending in .pdf or .psd
  let file_names = [
    "CAT-pic-912.png", // OK
    "dog-pic-234.psd",
    "dOg-photo-876.png", // OK
    "rabbit-pic-194.jpg",
    "cat-pic-787.pdf",
    "cats-image-873.webp", // OK
    "cat-pic-090.jpg", // OK
  ];
  let rules = bounds_builder()
    .or_starting_with_ci(&["cat", "dog"])
    .and_not_ending_with_ci(&[".psd", ".pdf"]);
  let matched_files = file_names.filter_all_rules(&rules);
  /// This should yield ["CAT-pic-912.png", "dOg-photo-876.png", "cats-image-873.webp", "cat-pic-090.jpg"]

Version 0.3.17 Added new to_start(separator: &str) and to_remainder_start(separator: &str) method

This version introduced a missing to_start(&str) method to return the start of a string before the last occurrence of the separator, i.e. to opposite to_tail(). to_remainder_start() supplements this method returning start of a string before the last non-final occurrence of a separator, e.g. with the string one/two/three/, to_remainder_start("/") yields one/two, while to_start("/") yields one/two/three, removing only the trailing slash.

Version 0.3.16 expands the range of rules available

Introduced equals_ci(pattern: &str) and equals_ci_alphanum(pattern: &str) to match whole strings in case-sensitive with or with without non-alphanumeric characters stripped.

Version 0.3.15 expands the range of rules available

Corrected bug in is_numeric() when used with empty strings.

Version 0.3.0 expands the range of rules available

This version introduced a radical revision to the StringBounds enum with supplementary BoundsPosition and CaseMatchMode enums, to handle the full range of rules available via bounds_builder(). These rule sets may be used with with the matched_by_rules(), filter_all_rules() and filter_any_rules(). Full documentation for the 0.2.* series is available in the Github repo in the *v0-2* branch.

Version 0.2.5 introduces SimpleMatchAny and Whole matches in StringBounds.

This supplements SimpleMatchAll to apply or logic with rules sets (StringBound, tuples or simple strs). The StringBounds enum now has whole string match options (with case-insensitive and case-sensitive variants) to accommodate a mix of partial and whole string matches. It also adds a range of single-argument methods for boundsbuilder(). Versions of the _string-patterns crate before 0.3.0 contained many of these extensions. Since version 0.3.0 all traits, enums and methods defined in this simple-string-patterns have been removed. These crates supplement each other, but may be installed independently.

Version 0.2.2 introduces three new features:

• bounds_builder() makes it easier to define string matching rules methods requiring an array of StringBounds rules such as filter_all_conditional(). See example above.
• ToSegmentFromChars provides new methods to split on any of an array of characters, e.g. when processing common patterns that may use a predictable set of separators. This mimics characters classes in regular expressions and is more efficient when you only need to allow for a limited set of split characters.
• MatchOccurrences has a variant find_char_indices method that accepts a char rather than a &str. This avoids any need to cast a character to a string.