Trimming and Translating

Trimming is exposed in C++ on the view: strip(byteset), lstrip(byteset), and rstrip(byteset) return a narrowed view with the matching leading/trailing bytes removed — no copy is made.

#include <stringzilla/stringzilla.hpp>
namespace sz = ashvardanian::stringzilla;

int main() {
    sz::string_view_t padded = "  \t spaced out \n";
    sz::string_view_t tight = padded.strip(sz::whitespaces_set());
    assert(tight == "spaced out");
    return 0;
}

Translation — mapping every byte through a 256-entry lookup table — is offered both at the C ABI level through sz_strings_lookup_best, documented under String Collections and Containers, and in C++ through sz::lookup:

#include <stringzilla/stringzilla.hpp>
namespace sz = ashvardanian::stringzilla;

int main() {
    sz::look_up_table_t to_upper = sz::look_up_table_t::identity();
    for (int i = 'a'; i <= 'z'; ++i) to_upper[(char)i] = (char)(i - 32);

    sz::string_t text;
    if (sz::failed(text.assign("hello"))) return 1;
    text.lookup(to_upper); // in place
    assert(text == "HELLO");
    return 0;
}

The C side ships ready-made table initializers — sz_strings_lookup_init_lower, sz_strings_lookup_init_upper, and sz_strings_lookup_init_ascii. To select an ASCII-only fast path, scan for the complement of sz_byteset_init_ascii with sz_find_byteset_best, which reaches the SIMD byteset kernels instead of a scalar loop.