Expose the DocumentId struct to be sure to inject the generated ids

milli/src/update/index_documents/enrich.rs

@@ -2,6 +2,7 @@ use std::io::{Read, Seek};
 use std::result::Result as StdResult;
 use std::{fmt, iter};
 
+use serde::{Deserialize, Serialize};
 use serde_json::Value;
 
 use crate::documents::{DocumentsBatchIndex, DocumentsBatchReader, EnrichedDocumentsBatchReader};
@@ -89,14 +90,15 @@ pub fn enrich_documents_batch<R: Read + Seek>(
             Err(user_error) => return Ok(Err(user_error)),
         };
 
-        external_ids.insert(count.to_be_bytes(), document_id.value())?;
-
         if let Some(geo_value) = geo_field_id.and_then(|fid| document.get(fid)) {
             if let Err(user_error) = validate_geo_from_json(&document_id, geo_value)? {
                 return Ok(Err(UserError::from(user_error)));
             }
         }
 
+        let document_id = serde_json::to_vec(&document_id).map_err(InternalError::SerdeJson)?;
+        external_ids.insert(count.to_be_bytes(), document_id)?;
+
         count += 1;
     }
 
@@ -210,7 +212,7 @@ impl PrimaryKey<'_> {
 ///
 /// In case the document id has been auto-generated, the document nth is kept to help
 /// users debug if there is an issue with the document itself.
-#[derive(Clone)]
+#[derive(Serialize, Deserialize, Clone)]
 pub enum DocumentId {
     Retrieved { value: String },
     Generated { value: String, document_nth: u32 },
@@ -225,16 +227,20 @@ impl DocumentId {
         DocumentId::Generated { value, document_nth }
     }
 
-    fn value(&self) -> &str {
+    fn debug(&self) -> String {
+        format!("{:?}", self)
+    }
+
+    pub fn is_generated(&self) -> bool {
+        matches!(self, DocumentId::Generated { .. })
+    }
+
+    pub fn value(&self) -> &str {
         match self {
             DocumentId::Retrieved { value } => value,
             DocumentId::Generated { value, .. } => value,
         }
     }
-
-    fn debug(&self) -> String {
-        format!("{:?}", self)
-    }
 }
 
 impl fmt::Debug for DocumentId {

milli/src/update/index_documents/mod.rs

@@ -22,7 +22,7 @@ use typed_chunk::{write_typed_chunk_into_index, TypedChunk};
 use self::enrich::enrich_documents_batch;
 pub use self::enrich::{
     extract_float_from_value, validate_document_id, validate_document_id_value,
-    validate_geo_from_json,
+    validate_geo_from_json, DocumentId,
 };
 pub use self::helpers::{
     as_cloneable_grenad, create_sorter, create_writer, fst_stream_into_hashset,

milli/src/update/index_documents/transform.rs

@@ -153,8 +153,9 @@ impl<'a, 'i> Transform<'a, 'i> {
         let mapping = create_fields_mapping(&mut self.fields_ids_map, fields_index)?;
 
         let primary_key = cursor.primary_key().to_string();
-        self.fields_ids_map.insert(&primary_key).ok_or(UserError::AttributeLimitReached)?;
         let primary_key_id_nested = primary_key.contains('.');
+        let primary_key_id =
+            self.fields_ids_map.insert(&primary_key).ok_or(UserError::AttributeLimitReached)?;
 
         let mut flattened_document = None;
         let mut obkv_buffer = Vec::new();
@@ -162,7 +163,7 @@ impl<'a, 'i> Transform<'a, 'i> {
         let mut documents_count = 0;
         let mut field_buffer: Vec<(u16, Cow<[u8]>)> = Vec::new();
         while let Some(enriched_document) = cursor.next_enriched_document()? {
-            let EnrichedDocument { document, external_id } = enriched_document;
+            let EnrichedDocument { document, document_id } = enriched_document;
 
             let mut field_buffer_cache = drop_and_reuse(field_buffer);
             if self.indexer_settings.log_every_n.map_or(false, |len| documents_count % len == 0) {
@@ -171,6 +172,14 @@ impl<'a, 'i> Transform<'a, 'i> {
                 });
             }
 
+            // When the document id has been auto-generated by the `enrich_documents_batch`
+            // we must insert this document id into the remaped document.
+            let external_id = document_id.value();
+            if document_id.is_generated() {
+                let docid = serde_json::to_vec(external_id).map_err(InternalError::SerdeJson)?;
+                field_buffer_cache.push((primary_key_id, Cow::from(docid)));
+            }
+
             for (k, v) in document.iter() {
                 let mapped_id =
                     *mapping.get(&k).ok_or(InternalError::FieldIdMappingMissingEntry { key: k })?;

milli/src/update/mod.rs

@@ -3,7 +3,7 @@ pub use self::clear_documents::ClearDocuments;
 pub use self::delete_documents::{DeleteDocuments, DocumentDeletionResult};
 pub use self::facets::Facets;
 pub use self::index_documents::{
-    DocumentAdditionResult, IndexDocuments, IndexDocumentsConfig, IndexDocumentsMethod,
+    DocumentAdditionResult, DocumentId, IndexDocuments, IndexDocumentsConfig, IndexDocumentsMethod,
 };
 pub use self::indexer_config::IndexerConfig;
 pub use self::settings::{Setting, Settings};