WIP refactor data event pipeline
This commit is contained in:
@@ -99,6 +99,10 @@ impl LogItem {
|
||||
|
||||
pub type Sitemty<T> = Result<StreamItem<RangeCompletableItem<T>>, Error>;
|
||||
|
||||
pub fn sitem_data<X>(x: X) -> Sitemty<X> {
|
||||
Ok(StreamItem::DataItem(RangeCompletableItem::Data(x)))
|
||||
}
|
||||
|
||||
struct VisitLevel;
|
||||
|
||||
impl<'de> Visitor<'de> for VisitLevel {
|
||||
|
||||
@@ -1,6 +1,12 @@
|
||||
use crate::{RangeCompletableItem, Sitemty, StreamItem, WithLen};
|
||||
use err::Error;
|
||||
use futures_util::{Stream, StreamExt};
|
||||
use netpod::log::*;
|
||||
use serde::Serialize;
|
||||
use serde_json::Value as JsonValue;
|
||||
use std::fmt;
|
||||
use std::time::Duration;
|
||||
use tokio::time::timeout_at;
|
||||
|
||||
pub trait Collector: Send + Unpin + WithLen {
|
||||
type Input: Collectable;
|
||||
@@ -45,3 +51,98 @@ impl ToJsonResult for Sitemty<serde_json::Value> {
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
// TODO rename, it is also used for binned:
|
||||
pub async fn collect_plain_events_json<T, S>(
|
||||
stream: S,
|
||||
timeout: Duration,
|
||||
bin_count_exp: u32,
|
||||
events_max: u64,
|
||||
do_log: bool,
|
||||
) -> Result<JsonValue, Error>
|
||||
where
|
||||
S: Stream<Item = Sitemty<T>> + Unpin,
|
||||
T: Collectable + fmt::Debug,
|
||||
{
|
||||
let deadline = tokio::time::Instant::now() + timeout;
|
||||
// TODO in general a Collector does not need to know about the expected number of bins.
|
||||
// It would make more sense for some specific Collector kind to know.
|
||||
// Therefore introduce finer grained types.
|
||||
let mut collector = <T as Collectable>::new_collector(bin_count_exp);
|
||||
let mut i1 = 0;
|
||||
let mut stream = stream;
|
||||
let mut total_duration = Duration::ZERO;
|
||||
loop {
|
||||
let item = if i1 == 0 {
|
||||
stream.next().await
|
||||
} else {
|
||||
if false {
|
||||
None
|
||||
} else {
|
||||
match timeout_at(deadline, stream.next()).await {
|
||||
Ok(k) => k,
|
||||
Err(_) => {
|
||||
collector.set_timed_out();
|
||||
None
|
||||
}
|
||||
}
|
||||
}
|
||||
};
|
||||
match item {
|
||||
Some(item) => {
|
||||
match item {
|
||||
Ok(item) => match item {
|
||||
StreamItem::Log(item) => {
|
||||
if do_log {
|
||||
debug!("collect_plain_events_json log {:?}", item);
|
||||
}
|
||||
}
|
||||
StreamItem::Stats(item) => {
|
||||
use crate::StatsItem;
|
||||
use netpod::DiskStats;
|
||||
match item {
|
||||
// TODO factor and simplify the stats collection:
|
||||
StatsItem::EventDataReadStats(_) => {}
|
||||
StatsItem::RangeFilterStats(_) => {}
|
||||
StatsItem::DiskStats(item) => match item {
|
||||
DiskStats::OpenStats(k) => {
|
||||
total_duration += k.duration;
|
||||
}
|
||||
DiskStats::SeekStats(k) => {
|
||||
total_duration += k.duration;
|
||||
}
|
||||
DiskStats::ReadStats(k) => {
|
||||
total_duration += k.duration;
|
||||
}
|
||||
DiskStats::ReadExactStats(k) => {
|
||||
total_duration += k.duration;
|
||||
}
|
||||
},
|
||||
}
|
||||
}
|
||||
StreamItem::DataItem(item) => match item {
|
||||
RangeCompletableItem::RangeComplete => {
|
||||
collector.set_range_complete();
|
||||
}
|
||||
RangeCompletableItem::Data(item) => {
|
||||
collector.ingest(&item);
|
||||
i1 += 1;
|
||||
if i1 >= events_max {
|
||||
break;
|
||||
}
|
||||
}
|
||||
},
|
||||
},
|
||||
Err(e) => {
|
||||
// TODO Need to use some flags to get good enough error message for remote user.
|
||||
Err(e)?;
|
||||
}
|
||||
};
|
||||
}
|
||||
None => break,
|
||||
}
|
||||
}
|
||||
let ret = serde_json::to_value(collector.result()?)?;
|
||||
debug!("Total duration: {:?}", total_duration);
|
||||
Ok(ret)
|
||||
}
|
||||
|
||||
Reference in New Issue
Block a user