Description
A fast, lightweight and single-header csv parser for modern C++
lazycsv alternatives and similar libraries
Based on the "CSV" category.
Alternatively, view lazycsv alternatives based on common mentions on social networks and blogs.
-
Vince's CSV Parser
A modern C++ library for reading, writing, and analyzing CSV (and similar) files.
WorkOS - The modern identity platform for B2B SaaS
Do you think we are missing an alternative of lazycsv or a related project?
README
[lazycsv](img/logo.png)
What's the lazycsv?
lazycsv is a c++17, posix-compliant, single-header library for reading and parsing csv files.
It's fast and lightweight and does not allocate any memory in the constructor or while parsing. It parses each row and cell just on demand on each iteration, that's why it's called lazy.
Note
This parser does not handle quoted cells yet.
Quick usage
The latest version of the single header can be downloaded from [include/lazycsv.hpp](include/lazycsv.hpp).
#include <lazycsv.hpp>
int main()
{
lazycsv::parser parser{ "contacts.csv" };
for (const auto row : parser)
{
const auto [id, name, phone] = row.cells(0, 1, 4); // indexes must be in ascending order
}
}
Performance note
Parser doesn't keep state of already parsed rows and cells, iterating through them always associated with parsing cost. This is true with cells() member function too, geting all needed cells by a single call is recommended.
If it's necessary to return to the already parsed rows and cells, they can be stored in a container and used later without being parsed again (they are view objects and efficient to copy).
Features
Returned std::string_view by raw() and trimed() member functions are valid as long as the parser object is alive:
std::vector<std::string_view> cities;
for (const auto row : parser)
{
const auto [city, state] = row.cells(0, 1);
cities.push_back(city.trimed());
}
Iterate through rows and cells:
for (const auto row : parser)
{
for (const auto cell : row)
{
}
}
Get header row and iterate through its cells:
auto header = parser.header();
for (const auto cell : header)
{
}
Find column index by its name:
auto city_index = parser.index_of("city");
row and cell are view objects on actual data in the parser object, they can be stored and used as long as the parser object is alive:
std::vector<lazycsv::parser<>::row> desired_rows;
std::vector<lazycsv::parser<>::cell> desired_cells;
for (const auto row : parser)
{
const auto [city] = row.cells(6);
desired_cells.push_back(city);
if (city.trimed() == "Kashan")
desired_rows.push_back(row);
}
static_assert(sizeof(lazycsv::parser<>::row) == 2 * sizeof(void*)); // i'm lightweight
static_assert(sizeof(lazycsv::parser<>::cell) == 2 * sizeof(void*)); // i'm lightweight too
Parser is customizable with the template parameters:
lazycsv::parser<
lazycsv::mmap_source, /* source type of csv data */
lazycsv::has_header<true>, /* first row is header or not */
lazycsv::delimiter<','>, /* column delimiter */
lazycsv::trim_chars<' ', '\t'>> /* trim characters of cells */
my_parser{ "data.csv" };
By default parser uses lazycsv::mmap_source as its source of data, but it's possible to be used with any other types of contiguous containers:
std::string csv_data{ "name,lastname,age\nPeter,Griffin,45\nchris,Griffin,14\n" };
lazycsv::parser<std::string_view> parser_a{ csv_data };
lazycsv::parser<std::string> parser_b{ csv_data };
TODO
- Add benchmarks
- Add parser for integral and floating point cells
- Add policy for quoting
- Wrapping it for python