# iterator.erl
> Lazy sequences simulating stdlib `lists` module API
[](https://github.com/klarna-incubator/iterator-erl/actions/workflows/ci.yml)
[![License][license-image]][license-url]
[![Developed at Klarna][klarna-image]][klarna-url]
This library provides an interface to define your own "lazy sequences" and also provides a list of
helpers to operate on such sequences that mimicks the OTP stdlib [lists](https://www.erlang.org/doc/man/lists)
functions.
"lazy sequences" allows you to operate on potentially huge streams of data the same way you would
operate on them if they are completely loaded into memory as a very long list, but doing so only
using constant memory footprint. It achieves that by loading one-list-item at a time.
## Usage example
It's primary intention is that you could define your own "iterators" using our generic interface
and then use our helper functions to build the "processing" pipeline for the stream of data.
For example, let's implement an iterator that returns next line from a file each time:
```erlang
file_line_iterator(FileName) ->
{ok, Fd} = file:open(Filename, [read, read_ahead, raw]),
iterator:new(fun yield_file_line/1, Fd, fun close_file/1).
yield_file_line(Fd) ->
case file:read_line(Fd) of
{ok, Line} ->
{Line, Fd};
eof ->
done;
{error, Reason} ->
error(Reason)
end.
close_file(Fd) ->
file:close(Fd).
```
You use `iterator:new/3` to create the (opaque) iterator structure, where
* 1st argument is the "yield" function - the function that returns either
`{NextSequenceElement, NewState}` or atom `done` when sequence is exhausted
* 2nd argument is the initial state of the iterator (in this example it does not change)
* 3rd optional argument is the "close" garbage-collection function that will be called when
iterator is exhausted or discarded (eg, used with `iterator:takewhile/2`)
After that you can use the primitive `iterator:next/1` function that returns either
`{ok, NextElement, NewIterator}` or `done`. But the real power comes when you build
a "processing pipeline" instead.
Then we may build a "processing pipeline" for this iterator. Let's say we want to filter-out the
lines that match a regular expression:
```erlang
LinesIterator = file_line_iterator("my_file.txt"),
MatchingIterator =
iterator:filter(
fun(Line) ->
case re:run(Line, "^[0-9]+$") of
nomatch ->
false;
{match, _} ->
true
end
end, LinesIterator).
```
And then we want to convert each matching line to integer
```erlang
IntegerIterator = iterator:map(fun erlang:binary_to_integer/1, MatchingIterator).
```
And finally we want to sum all the integers
```erlang
Sum = iterator:fold(fun(Int, Acc) -> Int + Acc end, 0, IntegerIterator).
```
The `iterator:fold/2` is different from other pipeline functions because it does not return
the new iterator, but it "forces" the execution of iterator by reading inner iterator's elements
one-by-one and applying `fun` to them, maintaining the `Acc` state.
Another such functions are `iterator:to_list/1`, `iterator:search/2`, `iterator:mapfoldl/3`.
With this code, using `iterator`, we managed to go through the whole file never keeping more than
a single line in-memory but were able to work with it using the same code style and high-order
functions as what we would use if we read all the file lines in memory.
## Setup
Add it to your `rebar.config`
```erlang
{deps, [iterator]}.
```
## Release History
See our [changelog](CHANGELOG.md).
## License
Copyright © 2023 Klarna Bank AB
For license details, see the [LICENSE](LICENSE) file in the root of this project.
<!-- Markdown link & img dfn's -->
[ci-image]: https://img.shields.io/badge/build-passing-brightgreen?style=flat-square
[ci-url]: https://github.com/klarna-incubator/TODO
[license-image]: https://img.shields.io/badge/license-Apache%202-blue?style=flat-square
[license-url]: http://www.apache.org/licenses/LICENSE-2.0
[klarna-image]: https://img.shields.io/badge/%20-Developed%20at%20Klarna-black?style=flat-square&labelColor=ffb3c7&logo=klarna&logoColor=black
[klarna-url]: https://klarna.github.io
