Merge #3164

3164: Improve the way we receive the documents payload r=Kerollmops a=jiangbo212

# Pull Request

## Related issue
Fixes #3037 

## What does this PR do?
- writing the playload to a temporary file via BufWritter
- deserialising the json tempporary file to an array of Objects by means of a memory map
- deserialising thie csv tempporary file by means of a memory map
- Adapted some read_json tests

## PR checklist
Please check if your PR fulfills the following requirements:
- [x] Does this PR fix an existing issue, or have you listed the changes applied in the PR description (and why they are needed)?
- [x] Have you read the contributing guidelines?
- [x] Have you made sure that the title is accurate and descriptive of the changes?

Thank you so much for contributing to Meilisearch!


Co-authored-by: jiangbo212 <peiyaoliukuan@gmail.com>
Co-authored-by: jiangbo212 <peiyaoliukuan@126.com>

Cargo.lock

@@ -2378,6 +2378,7 @@ dependencies = [
  "fst",
  "insta",
  "meili-snap",
+ "memmap2",
  "milli",
  "proptest",
  "proptest-derive",

index-scheduler/src/lib.rs

@@ -1065,19 +1065,21 @@ impl IndexScheduler {
 
 #[cfg(test)]
 mod tests {
+    use std::io::{BufWriter, Seek, Write};
     use std::time::Instant;
 
     use big_s::S;
     use crossbeam::channel::RecvTimeoutError;
     use file_store::File;
     use meili_snap::snapshot;
+    use meilisearch_types::document_formats::DocumentFormatError;
     use meilisearch_types::milli::obkv_to_json;
     use meilisearch_types::milli::update::IndexDocumentsMethod::{
         ReplaceDocuments, UpdateDocuments,
     };
     use meilisearch_types::tasks::IndexSwap;
     use meilisearch_types::VERSION_FILE_NAME;
-    use tempfile::TempDir;
+    use tempfile::{NamedTempFile, TempDir};
     use time::Duration;
     use uuid::Uuid;
     use Breakpoint::*;
@@ -1184,6 +1186,18 @@ mod tests {
         }
     }
 
+    /// Adapting to the new json reading interface
+    pub fn read_json(
+        bytes: &[u8],
+        write: impl Write + Seek,
+    ) -> std::result::Result<usize, DocumentFormatError> {
+        let temp_file = NamedTempFile::new().unwrap();
+        let mut buffer = BufWriter::new(temp_file.reopen().unwrap());
+        buffer.write_all(bytes).unwrap();
+        buffer.flush().unwrap();
+        meilisearch_types::document_formats::read_json(temp_file.as_file(), write)
+    }
+
     /// Create an update file with the given file uuid.
     ///
     /// The update file contains just one simple document whose id is given by `document_id`.
@@ -1202,9 +1216,7 @@ mod tests {
         );
 
         let (_uuid, mut file) = index_scheduler.create_update_file_with_uuid(file_uuid).unwrap();
-        let documents_count =
-            meilisearch_types::document_formats::read_json(content.as_bytes(), file.as_file_mut())
-                .unwrap() as u64;
+        let documents_count = read_json(content.as_bytes(), file.as_file_mut()).unwrap() as u64;
         (file, documents_count)
     }
 
@@ -1584,9 +1596,7 @@ mod tests {
         }"#;
 
         let (uuid, mut file) = index_scheduler.create_update_file_with_uuid(0).unwrap();
-        let documents_count =
-            meilisearch_types::document_formats::read_json(content.as_bytes(), file.as_file_mut())
-                .unwrap() as u64;
+        let documents_count = read_json(content.as_bytes(), file.as_file_mut()).unwrap() as u64;
         file.persist().unwrap();
         index_scheduler
             .register(KindWithContent::DocumentAdditionOrUpdate {
@@ -1623,9 +1633,7 @@ mod tests {
         snapshot!(snapshot_index_scheduler(&index_scheduler), name: "registered_the_first_task");
 
         let (uuid, mut file) = index_scheduler.create_update_file_with_uuid(0).unwrap();
-        let documents_count =
-            meilisearch_types::document_formats::read_json(content.as_bytes(), file.as_file_mut())
-                .unwrap() as u64;
+        let documents_count = read_json(content.as_bytes(), file.as_file_mut()).unwrap() as u64;
         file.persist().unwrap();
         index_scheduler
             .register(KindWithContent::DocumentAdditionOrUpdate {
@@ -1792,9 +1800,7 @@ mod tests {
         }"#;
 
         let (uuid, mut file) = index_scheduler.create_update_file_with_uuid(0).unwrap();
-        let documents_count =
-            meilisearch_types::document_formats::read_json(content.as_bytes(), file.as_file_mut())
-                .unwrap() as u64;
+        let documents_count = read_json(content.as_bytes(), file.as_file_mut()).unwrap() as u64;
         file.persist().unwrap();
         index_scheduler
             .register(KindWithContent::DocumentAdditionOrUpdate {
@@ -1952,11 +1958,7 @@ mod tests {
             );
 
             let (uuid, mut file) = index_scheduler.create_update_file_with_uuid(i).unwrap();
-            let documents_count = meilisearch_types::document_formats::read_json(
-                content.as_bytes(),
-                file.as_file_mut(),
-            )
-            .unwrap() as u64;
+            let documents_count = read_json(content.as_bytes(), file.as_file_mut()).unwrap() as u64;
             file.persist().unwrap();
             index_scheduler
                 .register(KindWithContent::DocumentAdditionOrUpdate {
@@ -2003,11 +2005,7 @@ mod tests {
             );
 
             let (uuid, mut file) = index_scheduler.create_update_file_with_uuid(i).unwrap();
-            let documents_count = meilisearch_types::document_formats::read_json(
-                content.as_bytes(),
-                file.as_file_mut(),
-            )
-            .unwrap() as u64;
+            let documents_count = read_json(content.as_bytes(), file.as_file_mut()).unwrap() as u64;
             file.persist().unwrap();
             index_scheduler
                 .register(KindWithContent::DocumentAdditionOrUpdate {
@@ -2056,11 +2054,7 @@ mod tests {
             );
 
             let (uuid, mut file) = index_scheduler.create_update_file_with_uuid(i).unwrap();
-            let documents_count = meilisearch_types::document_formats::read_json(
-                content.as_bytes(),
-                file.as_file_mut(),
-            )
-            .unwrap() as u64;
+            let documents_count = read_json(content.as_bytes(), file.as_file_mut()).unwrap() as u64;
             file.persist().unwrap();
             index_scheduler
                 .register(KindWithContent::DocumentAdditionOrUpdate {
@@ -2110,11 +2104,7 @@ mod tests {
             );
 
             let (uuid, mut file) = index_scheduler.create_update_file_with_uuid(i).unwrap();
-            let documents_count = meilisearch_types::document_formats::read_json(
-                content.as_bytes(),
-                file.as_file_mut(),
-            )
-            .unwrap() as u64;
+            let documents_count = read_json(content.as_bytes(), file.as_file_mut()).unwrap() as u64;
             file.persist().unwrap();
             index_scheduler
                 .register(KindWithContent::DocumentAdditionOrUpdate {
@@ -2165,11 +2155,7 @@ mod tests {
             );
 
             let (uuid, mut file) = index_scheduler.create_update_file_with_uuid(i).unwrap();
-            let documents_count = meilisearch_types::document_formats::read_json(
-                content.as_bytes(),
-                file.as_file_mut(),
-            )
-            .unwrap() as u64;
+            let documents_count = read_json(content.as_bytes(), file.as_file_mut()).unwrap() as u64;
             file.persist().unwrap();
             index_scheduler
                 .register(KindWithContent::DocumentAdditionOrUpdate {
@@ -2616,9 +2602,7 @@ mod tests {
         }"#;
 
         let (uuid, mut file) = index_scheduler.create_update_file_with_uuid(0).unwrap();
-        let documents_count =
-            meilisearch_types::document_formats::read_json(content.as_bytes(), file.as_file_mut())
-                .unwrap() as u64;
+        let documents_count = read_json(content.as_bytes(), file.as_file_mut()).unwrap() as u64;
         file.persist().unwrap();
         index_scheduler
             .register(KindWithContent::DocumentAdditionOrUpdate {
@@ -2656,9 +2640,7 @@ mod tests {
         }"#;
 
         let (uuid, mut file) = index_scheduler.create_update_file_with_uuid(0).unwrap();
-        let documents_count =
-            meilisearch_types::document_formats::read_json(content.as_bytes(), file.as_file_mut())
-                .unwrap() as u64;
+        let documents_count = read_json(content.as_bytes(), file.as_file_mut()).unwrap() as u64;
         file.persist().unwrap();
         index_scheduler
             .register(KindWithContent::DocumentAdditionOrUpdate {
@@ -2714,11 +2696,7 @@ mod tests {
             );
 
             let (uuid, mut file) = index_scheduler.create_update_file_with_uuid(i).unwrap();
-            let documents_count = meilisearch_types::document_formats::read_json(
-                content.as_bytes(),
-                file.as_file_mut(),
-            )
-            .unwrap() as u64;
+            let documents_count = read_json(content.as_bytes(), file.as_file_mut()).unwrap() as u64;
             file.persist().unwrap();
             index_scheduler
                 .register(KindWithContent::DocumentAdditionOrUpdate {
@@ -2766,11 +2744,7 @@ mod tests {
             );
 
             let (uuid, mut file) = index_scheduler.create_update_file_with_uuid(i).unwrap();
-            let documents_count = meilisearch_types::document_formats::read_json(
-                content.as_bytes(),
-                file.as_file_mut(),
-            )
-            .unwrap() as u64;
+            let documents_count = read_json(content.as_bytes(), file.as_file_mut()).unwrap() as u64;
             file.persist().unwrap();
             index_scheduler
                 .register(KindWithContent::DocumentAdditionOrUpdate {
@@ -2824,11 +2798,7 @@ mod tests {
             );
 
             let (uuid, mut file) = index_scheduler.create_update_file_with_uuid(i).unwrap();
-            let documents_count = meilisearch_types::document_formats::read_json(
-                content.as_bytes(),
-                file.as_file_mut(),
-            )
-            .unwrap() as u64;
+            let documents_count = read_json(content.as_bytes(), file.as_file_mut()).unwrap() as u64;
             file.persist().unwrap();
             index_scheduler
                 .register(KindWithContent::DocumentAdditionOrUpdate {
@@ -2887,11 +2857,7 @@ mod tests {
             );
 
             let (uuid, mut file) = index_scheduler.create_update_file_with_uuid(i).unwrap();
-            let documents_count = meilisearch_types::document_formats::read_json(
-                content.as_bytes(),
-                file.as_file_mut(),
-            )
-            .unwrap() as u64;
+            let documents_count = read_json(content.as_bytes(), file.as_file_mut()).unwrap() as u64;
             file.persist().unwrap();
             index_scheduler
                 .register(KindWithContent::DocumentAdditionOrUpdate {
@@ -2955,11 +2921,7 @@ mod tests {
             let allow_index_creation = i % 2 != 0;
 
             let (uuid, mut file) = index_scheduler.create_update_file_with_uuid(i).unwrap();
-            let documents_count = meilisearch_types::document_formats::read_json(
-                content.as_bytes(),
-                file.as_file_mut(),
-            )
-            .unwrap() as u64;
+            let documents_count = read_json(content.as_bytes(), file.as_file_mut()).unwrap() as u64;
             file.persist().unwrap();
             index_scheduler
                 .register(KindWithContent::DocumentAdditionOrUpdate {
@@ -3012,11 +2974,7 @@ mod tests {
             let allow_index_creation = i % 2 != 0;
 
             let (uuid, mut file) = index_scheduler.create_update_file_with_uuid(i).unwrap();
-            let documents_count = meilisearch_types::document_formats::read_json(
-                content.as_bytes(),
-                file.as_file_mut(),
-            )
-            .unwrap() as u64;
+            let documents_count = read_json(content.as_bytes(), file.as_file_mut()).unwrap() as u64;
             file.persist().unwrap();
             index_scheduler
                 .register(KindWithContent::DocumentAdditionOrUpdate {

meilisearch-types/Cargo.toml

@@ -12,6 +12,7 @@ either = { version = "1.6.1", features = ["serde"] }
 enum-iterator = "1.1.3"
 flate2 = "1.0.24"
 fst = "0.4.7"
+memmap2 = "0.5.7"
 milli = { git = "https://github.com/meilisearch/milli.git", tag = "v0.37.2", default-features = false }
 proptest = { version = "1.0.0", optional = true }
 proptest-derive = { version = "0.3.0", optional = true }

meilisearch-types/src/document_formats.rs

@@ -1,11 +1,15 @@
 use std::borrow::Borrow;
 use std::fmt::{self, Debug, Display};
-use std::io::{self, BufReader, Read, Seek, Write};
+use std::fs::File;
+use std::io::{self, Seek, Write};
+use std::marker::PhantomData;
 
 use either::Either;
+use memmap2::MmapOptions;
 use milli::documents::{DocumentsBatchBuilder, Error};
 use milli::Object;
-use serde::Deserialize;
+use serde::de::{SeqAccess, Visitor};
+use serde::{Deserialize, Deserializer};
 use serde_json::error::Category;
 
 use crate::error::{Code, ErrorCode};
@@ -99,10 +103,10 @@ impl ErrorCode for DocumentFormatError {
 internal_error!(DocumentFormatError: io::Error);
 
 /// Reads CSV from input and write an obkv batch to writer.
-pub fn read_csv(input: impl Read, writer: impl Write + Seek) -> Result<usize> {
+pub fn read_csv(file: &File, writer: impl Write + Seek) -> Result<usize> {
     let mut builder = DocumentsBatchBuilder::new(writer);
-
-    let csv = csv::Reader::from_reader(input);
+    let mmap = unsafe { MmapOptions::new().map(file)? };
+    let csv = csv::Reader::from_reader(mmap.as_ref());
     builder.append_csv(csv).map_err(|e| (PayloadType::Csv, e))?;
 
     let count = builder.documents_count();
@@ -111,30 +115,38 @@ pub fn read_csv(input: impl Read, writer: impl Write + Seek) -> Result<usize> {
     Ok(count as usize)
 }
 
-/// Reads JSON Lines from input and write an obkv batch to writer.
-pub fn read_ndjson(input: impl Read, writer: impl Write + Seek) -> Result<usize> {
-    let mut builder = DocumentsBatchBuilder::new(writer);
-    let reader = BufReader::new(input);
-
-    for result in serde_json::Deserializer::from_reader(reader).into_iter() {
-        let object = result.map_err(Error::Json).map_err(|e| (PayloadType::Ndjson, e))?;
-        builder
-            .append_json_object(&object)
-            .map_err(Into::into)
-            .map_err(DocumentFormatError::Internal)?;
+/// Reads JSON from temporary file  and write an obkv batch to writer.
+pub fn read_json(file: &File, writer: impl Write + Seek) -> Result<usize> {
+    read_json_inner(file, writer, PayloadType::Json)
 }
 
-    let count = builder.documents_count();
-    let _ = builder.into_inner().map_err(Into::into).map_err(DocumentFormatError::Internal)?;
-
-    Ok(count as usize)
+/// Reads JSON from temporary file  and write an obkv batch to writer.
+pub fn read_ndjson(file: &File, writer: impl Write + Seek) -> Result<usize> {
+    read_json_inner(file, writer, PayloadType::Ndjson)
 }
 
-/// Reads JSON from input and write an obkv batch to writer.
-pub fn read_json(input: impl Read, writer: impl Write + Seek) -> Result<usize> {
+/// Reads JSON from temporary file  and write an obkv batch to writer.
+fn read_json_inner(
+    file: &File,
+    writer: impl Write + Seek,
+    payload_type: PayloadType,
+) -> Result<usize> {
     let mut builder = DocumentsBatchBuilder::new(writer);
-    let reader = BufReader::new(input);
+    let mmap = unsafe { MmapOptions::new().map(file)? };
+    let mut deserializer = serde_json::Deserializer::from_slice(&mmap);
 
+    match array_each(&mut deserializer, |obj: Object| builder.append_json_object(&obj)) {
+        // The json data has been successfully deserialised and does not need to be processed again.
+        // the data has been successfully transferred to the "update_file" during the deserialisation process.
+        // count ==0 means an empty array
+        Ok(Ok(count)) => {
+            if count == 0 {
+                return Ok(count as usize);
+            }
+        }
+        Ok(Err(e)) => return Err(DocumentFormatError::Internal(Box::new(e))),
+        // Prefer deserialization as a json array. Failure to do deserialisation using the traditional method.
+        Err(_e) => {
             #[derive(Deserialize, Debug)]
             #[serde(transparent)]
             struct ArrayOrSingleObject {
@@ -142,8 +154,9 @@ pub fn read_json(input: impl Read, writer: impl Write + Seek) -> Result<usize> {
                 inner: Either<Vec<Object>, Object>,
             }
 
-    let content: ArrayOrSingleObject =
-        serde_json::from_reader(reader).map_err(Error::Json).map_err(|e| (PayloadType::Json, e))?;
+            let content: ArrayOrSingleObject = serde_json::from_reader(file)
+                .map_err(Error::Json)
+                .map_err(|e| (payload_type, e))?;
 
             for object in content.inner.map_right(|o| vec![o]).into_inner() {
                 builder
@@ -151,9 +164,57 @@ pub fn read_json(input: impl Read, writer: impl Write + Seek) -> Result<usize> {
                     .map_err(Into::into)
                     .map_err(DocumentFormatError::Internal)?;
             }
+        }
+    }
 
     let count = builder.documents_count();
     let _ = builder.into_inner().map_err(Into::into).map_err(DocumentFormatError::Internal)?;
 
     Ok(count as usize)
 }
+
+/**
+ * The actual handling of the deserialization process in the serde avoids storing the deserialized object in memory.
+ * Reference:
+ * https://serde.rs/stream-array.html
+ * https://github.com/serde-rs/json/issues/160
+ */
+fn array_each<'de, D, T, F>(deserializer: D, f: F) -> std::result::Result<io::Result<u64>, D::Error>
+where
+    D: Deserializer<'de>,
+    T: Deserialize<'de>,
+    F: FnMut(T) -> io::Result<()>,
+{
+    struct SeqVisitor<T, F>(F, PhantomData<T>);
+
+    impl<'de, T, F> Visitor<'de> for SeqVisitor<T, F>
+    where
+        T: Deserialize<'de>,
+        F: FnMut(T) -> io::Result<()>,
+    {
+        type Value = io::Result<u64>;
+
+        fn expecting(&self, formatter: &mut fmt::Formatter) -> fmt::Result {
+            formatter.write_str("a nonempty sequence")
+        }
+
+        fn visit_seq<A>(
+            mut self,
+            mut seq: A,
+        ) -> std::result::Result<io::Result<u64>, <A as SeqAccess<'de>>::Error>
+        where
+            A: SeqAccess<'de>,
+        {
+            let mut max: u64 = 0;
+            while let Some(value) = seq.next_element::<T>()? {
+                match self.0(value) {
+                    Ok(()) => max += 1,
+                    Err(e) => return Ok(Err(e)),
+                };
+            }
+            Ok(Ok(max))
+        }
+    }
+    let visitor = SeqVisitor(f, PhantomData);
+    deserializer.deserialize_seq(visitor)
+}

meilisearch/src/error.rs

@@ -95,6 +95,8 @@ pub enum PayloadError {
     MalformedPayload(serde_json::error::Error),
     #[error("A json payload is missing.")]
     MissingPayload,
+    #[error("Error while writing the playload to disk: `{0}`.")]
+    ReceivePayloadErr(Box<dyn std::error::Error + Send + Sync + 'static>),
 }
 
 impl ErrorCode for PayloadError {
@@ -126,6 +128,7 @@ impl ErrorCode for PayloadError {
             },
             PayloadError::MissingPayload => Code::MissingPayload,
             PayloadError::MalformedPayload(_) => Code::MalformedPayload,
+            PayloadError::ReceivePayloadErr(_) => Code::Internal,
         }
     }
 }

meilisearch/src/routes/indexes/documents.rs

@@ -1,4 +1,4 @@
-use std::io::{Cursor, ErrorKind};
+use std::io::ErrorKind;
 
 use actix_web::http::header::CONTENT_TYPE;
 use actix_web::web::Data;
@@ -20,9 +20,13 @@ use once_cell::sync::Lazy;
 use serde::Deserialize;
 use serde_cs::vec::CS;
 use serde_json::Value;
+use tempfile::tempfile;
+use tokio::fs::File;
+use tokio::io::{AsyncSeekExt, AsyncWriteExt, BufWriter};
 
 use crate::analytics::{Analytics, DocumentDeletionKind};
 use crate::error::MeilisearchHttpError;
+use crate::error::PayloadError::ReceivePayloadErr;
 use crate::extractors::authentication::policies::*;
 use crate::extractors::authentication::GuardedData;
 use crate::extractors::payload::Payload;
@@ -227,26 +231,52 @@ async fn document_addition(
 
     let (uuid, mut update_file) = index_scheduler.create_update_file()?;
 
-    // TODO: this can be slow, maybe we should spawn a thread? But the payload isn't Send+Sync :weary:
-    // push the entire stream into a `Vec`.
-    // If someone sends us a never ending stream we're going to block the thread.
-    // TODO: Maybe we should write it to a file to reduce the RAM consumption
-    // and then reread it to convert it to obkv?
-    let mut buffer = Vec::new();
-    while let Some(bytes) = body.next().await {
-        buffer.extend_from_slice(&bytes?);
+    let temp_file = match tempfile() {
+        Ok(temp_file) => temp_file,
+        Err(e) => {
+            return Err(MeilisearchHttpError::Payload(ReceivePayloadErr(Box::new(e))));
         }
-    if buffer.is_empty() {
+    };
+
+    let async_file = File::from_std(temp_file);
+    let mut buffer = BufWriter::new(async_file);
+
+    let mut buffer_write_size: usize = 0;
+    while let Some(bytes) = body.next().await {
+        let byte = &bytes?;
+
+        if byte.is_empty() && buffer_write_size == 0 {
             return Err(MeilisearchHttpError::MissingPayload(format));
         }
-    let reader = Cursor::new(buffer);
+
+        match buffer.write_all(byte).await {
+            Ok(()) => buffer_write_size += 1,
+            Err(e) => {
+                return Err(MeilisearchHttpError::Payload(ReceivePayloadErr(Box::new(e))));
+            }
+        };
+    }
+
+    if let Err(e) = buffer.flush().await {
+        return Err(MeilisearchHttpError::Payload(ReceivePayloadErr(Box::new(e))));
+    }
+
+    if buffer_write_size == 0 {
+        return Err(MeilisearchHttpError::MissingPayload(format));
+    }
+
+    if let Err(e) = buffer.seek(std::io::SeekFrom::Start(0)).await {
+        return Err(MeilisearchHttpError::Payload(ReceivePayloadErr(Box::new(e))));
+    };
+
+    let read_file = buffer.into_inner().into_std().await;
 
     let documents_count =
         tokio::task::spawn_blocking(move || -> Result<_, MeilisearchHttpError> {
             let documents_count = match format {
-                PayloadType::Json => read_json(reader, update_file.as_file_mut())?,
-                PayloadType::Csv => read_csv(reader, update_file.as_file_mut())?,
-                PayloadType::Ndjson => read_ndjson(reader, update_file.as_file_mut())?,
+                PayloadType::Json => read_json(&read_file, update_file.as_file_mut())?,
+                PayloadType::Csv => read_csv(&read_file, update_file.as_file_mut())?,
+                PayloadType::Ndjson => read_ndjson(&read_file, update_file.as_file_mut())?,
             };
             // we NEED to persist the file here because we moved the `udpate_file` in another task.
             update_file.persist()?;

meilisearch/tests/documents/add_documents.rs

@@ -436,7 +436,7 @@ async fn error_add_malformed_ndjson_documents() {
     assert_eq!(
         response["message"],
         json!(
-            r#"The `ndjson` payload provided is malformed. `Couldn't serialize document value: key must be a string at line 2 column 2`."#
+            r#"The `ndjson` payload provided is malformed. `Couldn't serialize document value: trailing characters at line 2 column 1`."#
         )
     );
     assert_eq!(response["code"], json!("malformed_payload"));
@@ -456,7 +456,7 @@ async fn error_add_malformed_ndjson_documents() {
     assert_eq!(status_code, 400);
     assert_eq!(
         response["message"],
-        json!("The `ndjson` payload provided is malformed. `Couldn't serialize document value: key must be a string at line 2 column 2`.")
+        json!("The `ndjson` payload provided is malformed. `Couldn't serialize document value: trailing characters at line 2 column 1`.")
     );
     assert_eq!(response["code"], json!("malformed_payload"));
     assert_eq!(response["type"], json!("invalid_request"));