Trait RowStream

pub trait RowStream: IntoIterator<Item = RowResult> {
    // Required method
    fn headers(&self) -> &Headers;

    // Provided methods
    fn add(self, column: ColSpec) -> Result<Add<Self>>
       where Self: Sized { ... }
    fn del(self, columns: Vec<&str>) -> Del<'_, Self>
       where Self: Sized { ... }
    fn select(self, columns: Vec<&str>) -> Select<'_, Self>
       where Self: Sized { ... }
    fn add_with<F>(
        self,
        colname: &str,
        f: F,
    ) -> Result<AddWith<Self, F>, BuildError>
       where Self: Sized,
             F: FnMut(&Headers, &Row) -> Result<String> { ... }
    fn reduce(self, columns: Vec<Box<dyn Aggregate>>) -> Result<Reduce<Self>>
       where Self: Sized { ... }
    fn adjacent_group<F, R, G>(
        self,
        grouping: G,
        f: F,
    ) -> AdjacentGroup<Self, F, G>
       where F: Fn(MockStream<IntoIter<RowResult>>) -> R,
             R: RowStream,
             G: GroupCriteria,
             Self: Sized { ... }
    fn group<F, R, G>(self, grouping: G, f: F) -> Group<Self, F, G>
       where F: Fn(MockStream<IntoIter<RowResult>>) -> R,
             R: RowStream,
             G: GroupCriteria,
             Self: Sized { ... }
    fn flush<T>(self, target: T) -> Result<Flush<Self, T>>
       where Self: Sized { ... }
    fn review<F>(self, f: F) -> Inspect<Self, F>
       where Self: Sized,
             F: FnMut(&Headers, &RowResult) { ... }
    fn rename(self, old_name: &str, new_name: &str) -> Rename<Self>
       where Self: Sized { ... }
    fn map_row<F, H, R>(self, f: F, header_map: H) -> MapRow<Self, F>
       where Self: Sized,
             F: Fn(&Headers, &Row) -> Result<R>,
             H: Fn(&Headers) -> Headers,
             R: Iterator<Item = RowResult> { ... }
    fn map_col<F>(self, col: &str, f: F) -> MapCol<Self, F>
       where Self: Sized,
             F: Fn(&str) -> Result<String> { ... }
    fn filter_col<F>(self, col: &str, f: F) -> Result<FilterCol<Self, F>>
       where Self: Sized,
             F: Fn(&str) -> bool { ... }
    fn filter_row<F>(self, f: F) -> FilterRow<Self, F>
       where Self: Sized,
             F: Fn(&Headers, &Row) -> bool { ... }
}

This trait describes de behaviour of every component in the CSV transformation chain. Functions provided by this trait help construct the chain and can be chained.

Implement this trait to extend csvsc with your own processors.

Required Methods

fn headers(&self) -> &Headers

Must return headers as they are in this point of the chain. For example if implementor adds a column, its headers() function must return the new headers including the one just added.

Provided Methods

fn add(self, column: ColSpec) -> Result<Add<Self>>

where Self: Sized,

Add a column to each row of the stream.

New columns can be build arbitrarily from previous columns or from a specific column using a regular expression.

use csvsc::prelude::*;
use encoding::all::UTF_8;

let mut chain = InputStreamBuilder::from_paths(&["test/assets/1.csv"])
    .unwrap().build().unwrap()
    .add(
        Column::with_name("new column")
            .from_all_previous()
            .definition("{old col1} - {old col2}")
    ).unwrap();

See Column for options.

If you want to add a constant value or have some other requirement take a look at .add_with().

fn del(self, columns: Vec<&str>) -> Del<'_, Self>

where Self: Sized,

Deletes the specified columns from each row of the stream. If you have too many columns to delete perhaps instead use RowStream::select.

fn select(self, columns: Vec<&str>) -> Select<'_, Self>

where Self: Sized,

Outputs only the selected columns, ignoring the rest.

The returned rows contain its values in the order corresponding to the order in which the headers were given to this function. That means that this function can be used to reorder the headers.

If you only want do delete specific columns take a look at RowStream::del.

fn add_with<F>( self, colname: &str, f: F, ) -> Result<AddWith<Self, F>, BuildError>

where Self: Sized, F: FnMut(&Headers, &Row) -> Result<String>,

Adds a column to each row of the stream using a closure to compute its value.

This you can use to add a constant value also.

Example

use csvsc::prelude::*;
use encoding::all::UTF_8;

let mut chain = InputStreamBuilder::from_paths(&["test/assets/1.csv"])
    .unwrap().build().unwrap()
    .add_with("new col", |headers, row| {
        Ok("value".into())
    }).unwrap();

fn reduce(self, columns: Vec<Box<dyn Aggregate>>) -> Result<Reduce<Self>>

where Self: Sized,

Reduce all the incoming stream into one row, computing some aggregates in the way. All the stream collapses into one row.

The final row contains only the result of reducers and no other column but you might preserve a column using the .last() aggregate.

You’ll likely be using this inside a .group() or .adjacent_group().

Example

use csvsc::prelude::*;
use encoding::all::UTF_8;

let mut chain = InputStreamBuilder::from_paths(&["test/assets/chicken_north.csv"])
    .unwrap().build().unwrap()
    .group(["month"], |row_stream| {
        row_stream
            .reduce(vec![
                Reducer::with_name("avg").of_column("eggs per week").average().unwrap(),
            ]).unwrap()
    });

See Reducer for built-in aggregates.

fn adjacent_group<F, R, G>(self, grouping: G, f: F) -> AdjacentGroup<Self, F, G>

where F: Fn(MockStream<IntoIter<RowResult>>) -> R, R: RowStream, G: GroupCriteria, Self: Sized,

Groups rows by the given criteria, but assuming a “group” is a set of adjacent rows.

This means that sets of rows that meet the same criteria but are not adjacent will not be grouped together. Only use it if you are sure that your data follows this pattern and you want to take advantage of it.

An interesting advantage of using this is that only one group is kept in memory at a time.

See RowStream::group for more details.

Example

Consider a file test/assets/groups.csv with this contents. Notice that there are four adjacent groups that have the same value for column name: two with value a and two with b.

name,value
a,1
a,1
b,2
b,2
a,3
a,3
b,4
b,4

Then the following code works as expected, generating an average for all of the four adjacent groups that have the same value for column name.

use csvsc::prelude::*;

let mut rows = InputStreamBuilder::from_paths(&["test/assets/groups.csv"]).unwrap().build().unwrap()
    .adjacent_group(["name"], |row_stream| {
        row_stream
            .reduce(vec![
                Reducer::with_name("name").of_column("name").last("").unwrap(),
                Reducer::with_name("avg").of_column("value").average().unwrap(),
            ]).unwrap()
    })
    .into_iter();

assert_eq!(rows.next().unwrap().unwrap(), Row::from(vec!["a", "1"]));
assert_eq!(rows.next().unwrap().unwrap(), Row::from(vec!["b", "2"]));
assert_eq!(rows.next().unwrap().unwrap(), Row::from(vec!["a", "3"]));
assert_eq!(rows.next().unwrap().unwrap(), Row::from(vec!["b", "4"]));

fn group<F, R, G>(self, grouping: G, f: F) -> Group<Self, F, G>

where F: Fn(MockStream<IntoIter<RowResult>>) -> R, R: RowStream, G: GroupCriteria, Self: Sized,

Groups rows by the given criteria. You’ll be given a RowStream instance as the first argument of a closure that you can use to further process the grouped rows.

The first argument is the group criteria and it can be any of:

• A slice of &str: &["foo", "bar"],
• an array of &str: ["foo", "bar"],
• a closure Fn(&Headers, &Row) -> Hash,
• any type that implements GroupCriteria

In the first two cases the &strs are treated as column names. Rows having the same values for the specified columns will belong to the same group. Strings that don’t match any column name will be ignored.

In the closure case you’ll be given the headers and every row and you must return a hashable type that identifies the group where that row belongs.

GroupCriteria is a trait you can implement for your own types if you want to use them as grouping criteria.

Example

Consider the following file:

name,value
a,1
a,1
b,2
b,2
a,3
a,3
b,4
b,4

Then we can group for example using the column name and get the following results:

use csvsc::prelude::*;

let mut rows: Vec<_> = InputStreamBuilder::from_paths(&["test/assets/groups.csv"]).unwrap().build().unwrap()
    .group(["name"], |row_stream| {
        row_stream
            .reduce(vec![
                Reducer::with_name("name").of_column("name").last("").unwrap(),
                Reducer::with_name("avg").of_column("value").average().unwrap(),
            ]).unwrap()
    })
    .into_iter()
    .filter_map(|r| r.ok())
    .collect();

rows.sort_by_key(|row| row.get(0).unwrap().to_string());

assert_eq!(rows[0], Row::from(vec!["a", "2"]));
assert_eq!(rows[1], Row::from(vec!["b", "3"]));

Grouping by closure

If you decide that you need an arbitrary grouping criteria you can use a closure that returns a hashable type like this:

use csvsc::prelude::*;

let mut rows: Vec<_> = InputStreamBuilder::from_paths(&["test/assets/groups.csv"]).unwrap().build().unwrap()
    .group(|headers: &Headers, row: &Row| {
        headers.get_field(row, "name").unwrap().to_string()
    }, |row_stream| {
        row_stream
            .reduce(vec![
                Reducer::with_name("name").of_column("name").last("").unwrap(),
                Reducer::with_name("avg").of_column("value").average().unwrap(),
            ]).unwrap()
    })
    .into_iter()
    .filter_map(|r| r.ok())
    .collect();

rows.sort_by_key(|row| row.get(0).unwrap().to_string());

assert_eq!(rows[0], Row::from(vec!["a", "2"]));
assert_eq!(rows[1], Row::from(vec!["b", "3"]));

fn flush<T>(self, target: T) -> Result<Flush<Self, T>>

where Self: Sized,

When consumed, writes to destination specified by the column given in the first argument. Other than that this behaves like an id(x) function so you can specify more links in the chain and even more flushers.

fn review<F>(self, f: F) -> Inspect<Self, F>

where Self: Sized, F: FnMut(&Headers, &RowResult),

Mostly for debugging, calls a closure on each element. Behaves like the identity function on the stream returning each row untouched.

fn rename(self, old_name: &str, new_name: &str) -> Rename<Self>

where Self: Sized,

Renames some columns

fn map_row<F, H, R>(self, f: F, header_map: H) -> MapRow<Self, F>

where Self: Sized, F: Fn(&Headers, &Row) -> Result<R>, H: Fn(&Headers) -> Headers, R: Iterator<Item = RowResult>,

Apply a function to every row and use the return values to build the row stream.

This method accepts a closure that must return an iterator over RowResult values, this means that you can create more rows out of a single one.

You’re responsible of providing the new headers and for that you need to use the second closure, that maps the old headers to the new ones.

Example

use csvsc::prelude::*;
use encoding::all::UTF_8;

InputStreamBuilder::from_paths(&["test/assets/1.csv"])
    .unwrap().build().unwrap()
    .map_row(|_headers, row| {
        // Go creative here in the creation of your new row(s)
        Ok(vec![
            Ok(row.clone())
        ].into_iter())
    }, |old_headers| {
        // be responsible and provide proper headers from the old ones
        old_headers.clone()
    })
    .into_iter();

fn map_col<F>(self, col: &str, f: F) -> MapCol<Self, F>

where Self: Sized, F: Fn(&str) -> Result<String>,

Apply a function to a single column of the stream, this function dones’t fail if the column dones’t exist.

fn filter_col<F>(self, col: &str, f: F) -> Result<FilterCol<Self, F>>

where Self: Sized, F: Fn(&str) -> bool,

filter entire rows out depending on one column’s value and a condition, leaving errored rows untouched.

fn filter_row<F>(self, f: F) -> FilterRow<Self, F>

where Self: Sized, F: Fn(&Headers, &Row) -> bool,

filter entire rows out depending on one column’s value and a condition, leaving errored rows untouched.

Implementors

impl RowStream for InputStream

impl<I> RowStream for Add<I>

where I: RowStream,

impl<I> RowStream for MockStream<I>

where MockStream<I>: IntoIterator<Item = RowResult>,