Skip to content
osmem

Compatibility

A test that passes on a fake proves something only if the fake behaves like the real system where the test looks. This page is the list you need to judge that for osmem: what is implemented, what is approximated, and what fails on purpose.

The rule

Section titled “The rule”

Unsupported features return an error, usually 400 unsupported_operation_exception with a reason ending in “not supported by osmem”, never a plausible empty result. Response shapes, status codes and error types follow OpenSearch 2.x so that client libraries decode them normally. Every difference found so far is covered by a regression test.

Implemented

Section titled “Implemented”

Indices. Create with settings, mappings and aliases; delete with wildcards and _all; exists; get; _mapping get and put (type conflicts are rejected as on OpenSearch); _settings; _stats; _refresh, _flush, _forcemerge, _open, _close as no-ops; _analyze; _cat/indices, _cat/aliases, _cat/health, _cat/count; cluster health, settings, state; _nodes with the real address for sniffing clients. Composable and legacy index templates apply on creation, including auto-created indices.

Documents. _doc with auto ids, op_type, if_seq_no/if_primary_term, external versions; _create; _update with doc, doc_as_upsert, upsert, detect_noop; _source; _mget; _bulk with per-item status; _delete_by_query; _update_by_query without a script; _reindex.

Mapping. text (analyzer, search_analyzer, multi-fields), keyword (normalizer, ignore_above), every numeric type, boolean, date and date_nanos with named formats, Java patterns and epoch formats, geo_point, ip, object and nested, null_value, copy_to, index: false, enabled: false, dynamic: true/false/strict. Dynamic mapping follows OpenSearch: strings become text with a .keyword sub-field, ISO dates are detected, integers become long and decimals float; dotted keys expand into objects.

Queries. match (operator, minimum_should_match, fuzziness, zero_terms_query), match_phrase with slop, match_phrase_prefix, match_bool_prefix, multi_match (best_fields, most_fields, cross_fields, phrase, phrase_prefix, bool_prefix, field boosts, wildcard field names), term typed by the mapping, terms including terms lookup, range on numbers, dates with date math, format and time_zone, and strings, exists, prefix, wildcard, regexp, fuzzy, ids, bool with all four clauses and minimum_should_match, constant_score, dis_max, query_string and simple_query_string, geo_distance, geo_bounding_box, wrapper.

Search. from/size up to the 10,000 window, sort by field with order, missing, mode, unmapped_type and format, _score, _doc, _id, search_after with OpenSearch’s sentinel values for missing numbers and dates, scroll, point in time, _source filtering, fields, docvalue_fields, version, seq_no_primary_term, track_total_hits, track_scores, min_score, post_filter, collapse, highlight, _count, _msearch, filter_path, pretty, rest_total_hits_as_int, gzip request bodies.

Aggregations. terms (size, order by count, key or sub-aggregation, min_doc_count, missing, include/exclude), multi_terms, range, date_range, histogram, date_histogram (calendar and fixed intervals, time_zone, offset, format, extended_bounds, gap filling), filter, filters, missing, global, nested and reverse_nested as pass-through, composite with after, avg, sum, min, max, value_count, stats, extended_stats, cardinality, percentiles, percentile_ranks, top_hits, weighted_avg, median_absolute_deviation; pipelines cumulative_sum, derivative, bucket_sort, avg/sum/min/max/stats_bucket. Sub-aggregations nest.

Analysis. standard, simple, whitespace, keyword, stop, pattern and bleve’s language analyzers; custom analyzers from tokenizers (standard, whitespace, keyword, letter, pattern, ngram, edge_ngram, char_group, uax_url_email), token filters (lowercase, asciifolding, stop, ngram, edge_ngram, shingle, stemmer, snowball, porter_stem, truncate, length, unique, reverse, cjk_bigram, cjk_width) and char filters (html_strip, pattern_replace); normalizers. With Japanese support: kuromoji analyzer, kuromoji_tokenizer with modes, user dictionaries, kuromoji_baseform, kuromoji_part_of_speech, cjk_width, ja_stop, kuromoji_stemmer, kuromoji_readingform.

Approximated

Section titled “Approximated”
  • Scores. bleve’s BM25 is not Lucene’s. Rankings agree for ordinary queries; _score values and ties do not. Filters are excluded from scoring as on OpenSearch.
  • Nested documents are flattened: a nested query matches when the fields match anywhere in the array, like an object field.
  • Analyzers. Language analyzers use bleve’s stemmers and stop lists. Japanese segmentation comes from kagome with the IPA dictionary rather than kuromoji and can differ on unknown words; Korean and Chinese are CJK bigrams. Unsupported filters in a custom analyzer are skipped with a warning.
  • Exactness. cardinality and percentiles are exact, while OpenSearch approximates them; a test asserting an approximate value would differ.
  • Refresh. Writes are visible immediately. A test cannot observe the window between a write and a refresh.

Not supported

Section titled “Not supported”

Painless in any form (script, script_score, function_score functions, scripted updates, bucket_script, bucket_selector), suggesters, kNN and neural search, percolator, join fields, span queries, geo shapes, significant_terms and other statistical aggregations, ingest pipelines, security, snapshots, node and shard management beyond stubs, changing max_result_window.

Client notes

Section titled “Client notes”
  • opensearch-go v4 is tested in CI. go-elasticsearch v8 receives the X-Elastic-Product header it checks.
  • olivere/elastic sniffs /_nodes; osmem reports the real listen address there.
  • For clients that insist on an Elasticsearch 7 version string, set the cluster setting compatibility.override_main_response_version: true; GET / then reports 7.10.2.
  • Errors follow the {"error": {"type", "reason", "root_cause"}, "status"} shape; unknown URLs return 400 with a string error, wrong methods 405, as on OpenSearch.