Panic occured while committing writer with sort by text field.

[gy0801151351]

Which version of tantivy are you using? tantivy v0.21.1

To Reproduce

use tantivy::schema::{SchemaBuilder, FAST, STORED, STRING, TEXT};
use tantivy::{doc, IndexBuilder, IndexSettings, IndexSortByField, Order};

fn main() -> tantivy::Result<()> {
    let mut schema_builder = SchemaBuilder::new();
    let id = schema_builder.add_text_field("id", STRING | FAST | STORED);
    let name = schema_builder.add_text_field("name", TEXT | STORED);

    let index = IndexBuilder::new()
        .schema(schema_builder.build())
        .settings(IndexSettings {
            sort_by_field: Some(IndexSortByField {
                field: "id".to_string(),
                order: Order::Asc,
            }),
            ..Default::default()
        })
        .create_in_ram()?;

    let mut writer = index.writer(50_000_000)?;
    writer.add_document(doc! {
        id => "0001",
        name => "name1"
    })?;
    writer.add_document(doc! {
        id => "0002",
        name => "name2"
    })?;
    writer.commit()?;
    writer.wait_merging_threads()?;

    Ok(())
}

index out of bounds: the len is 0 but the index is 0stack backtrace:

   0: std::panicking::begin_panic_handler
             at /rustc/8ace7ea1f7cbba7b4f031e66c54ca237a0d65de6/library\std\src\panicking.rs:647
   1: core::panicking::panic_fmt
             at /rustc/8ace7ea1f7cbba7b4f031e66c54ca237a0d65de6/library\core\src\panicking.rs:72
   2: core::panicking::panic_bounds_check
             at /rustc/8ace7ea1f7cbba7b4f031e66c54ca237a0d65de6/library\core\src\panicking.rs:208
   3: core::slice::index::impl$2::index<u32>
             at /rustc/8ace7ea1f7cbba7b4f031e66c54ca237a0d65de6\library\core\src\slice\index.rs:255
   4: alloc::vec::impl$12::index<u32,usize,alloc::alloc::Global>
             at /rustc/8ace7ea1f7cbba7b4f031e66c54ca237a0d65de6\library\alloc\src\vec\mod.rs:2771
   5: tantivy::indexer::doc_id_mapping::DocIdMapping::get_new_doc_id
             at E:\Library_Store\Cargo\registry\src\mirrors.ustc.edu.cn-61ef6e0cd06fb9b8\tantivy-0.21.1\src\indexer\doc_id_mapping.rs:94
   6: tantivy::postings::recorder::impl$4::serialize::closure$0
             at E:\Library_Store\Cargo\registry\src\mirrors.ustc.edu.cn-61ef6e0cd06fb9b8\tantivy-0.21.1\src\postings\recorder.rs:131
   7: core::ops::function::impls::impl$4::call_once<tuple$<u32>,tantivy::postings::recorder::impl$4::serialize::closure_env$0>
             at /rustc/8ace7ea1f7cbba7b4f031e66c54ca237a0d65de6\library\core\src\ops\function.rs:305
   8: enum2$<core::option::Option<u32> >::map
             at /rustc/8ace7ea1f7cbba7b4f031e66c54ca237a0d65de6\library\core\src\option.rs:1072
   9: core::iter::adapters::map::impl$2::next<u32,tantivy::postings::recorder::VInt32Reader,tantivy::postings::recorder::impl$4::serialize::closure_env$0>
             at /rustc/8ace7ea1f7cbba7b4f031e66c54ca237a0d65de6\library\core\src\iter\adapters\map.rs:108
  10: alloc::vec::Vec<u32,alloc::alloc::Global>::extend_desugared<u32,alloc::alloc::Global,core::iter::adapters::map::Map<tantivy::postings::recorder::VInt32Reader,tantivy::postings::recorder::impl$4::serialize::closure_env$0> >
             at /rustc/8ace7ea1f7cbba7b4f031e66c54ca237a0d65de6\library\alloc\src\vec\mod.rs:2930
  11: alloc::vec::spec_extend::impl$0::spec_extend<u32,core::iter::adapters::map::Map<tantivy::postings::recorder::VInt32Reader,tantivy::postings::recorder::impl$4::serialize::closure_env$0>,alloc::alloc::Global>
             at /rustc/8ace7ea1f7cbba7b4f031e66c54ca237a0d65de6\library\alloc\src\vec\spec_extend.rs:17
  12: alloc::vec::impl$18::extend<u32,alloc::alloc::Global,core::iter::adapters::map::Map<tantivy::postings::recorder::VInt32Reader,tantivy::postings::recorder::impl$4::serialize::closure_env$0> >
             at /rustc/8ace7ea1f7cbba7b4f031e66c54ca237a0d65de6\library\alloc\src\vec\mod.rs:2904
  13: tantivy::postings::recorder::impl$4::serialize
             at E:\Library_Store\Cargo\registry\src\mirrors.ustc.edu.cn-61ef6e0cd06fb9b8\tantivy-0.21.1\src\postings\recorder.rs:129
  14: tantivy::postings::postings_writer::SpecializedPostingsWriter<tantivy::postings::recorder::DocIdRecorder>::serialize_one_term<tantivy::postings::recorder::DocIdRecorder>
             at E:\Library_Store\Cargo\registry\src\mirrors.ustc.edu.cn-61ef6e0cd06fb9b8\tantivy-0.21.1\src\postings\postings_writer.rs:175
  15: tantivy::postings::postings_writer::impl$2::serialize<tantivy::postings::recorder::DocIdRecorder>
             at E:\Library_Store\Cargo\registry\src\mirrors.ustc.edu.cn-61ef6e0cd06fb9b8\tantivy-0.21.1\src\postings\postings_writer.rs:214
  16: tantivy::postings::postings_writer::serialize_postings
             at E:\Library_Store\Cargo\registry\src\mirrors.ustc.edu.cn-61ef6e0cd06fb9b8\tantivy-0.21.1\src\postings\postings_writer.rs:66
  17: tantivy::indexer::segment_writer::remap_and_write
             at E:\Library_Store\Cargo\registry\src\mirrors.ustc.edu.cn-61ef6e0cd06fb9b8\tantivy-0.21.1\src\indexer\segment_writer.rs:402
  18: tantivy::indexer::segment_writer::SegmentWriter::finalize
             at E:\Library_Store\Cargo\registry\src\mirrors.ustc.edu.cn-61ef6e0cd06fb9b8\tantivy-0.21.1\src\indexer\segment_writer.rs:148
  19: tantivy::indexer::index_writer::index_documents
             at E:\Library_Store\Cargo\registry\src\mirrors.ustc.edu.cn-61ef6e0cd06fb9b8\tantivy-0.21.1\src\indexer\index_writer.rs:199
  20: tantivy::indexer::index_writer::impl$0::add_indexing_worker::closure$0
             at E:\Library_Store\Cargo\registry\src\mirrors.ustc.edu.cn-61ef6e0cd06fb9b8\tantivy-0.21.1\src\indexer\index_writer.rs:428
note: Some details are omitted, run with `RUST_BACKTRACE=full` for a verbose backtrace.
stack backtrace:
   0: std::panicking::begin_panic_handler
Error:              at ErrorInThread/rustc/8ace7ea1f7cbba7b4f031e66c54ca237a0d65de6/library\std\src\panicking.rs(:"647Any { .. }
")
  1: core::panicking::panic_fmt
             at /rustc/8ace7ea1f7cbba7b4f031e66c54ca237a0d65de6/library\core\src\panicking.rs:72

[PSeitz]

sort by is only supported on numerical fields. Why do you want to sort by text field?

[gy0801151351]

sort by is only supported on numerical fields. Why do you want to sort by text field?

Because my source data is from MongoDB. Its id field is ObjectId type, and there is no way to store ObjectId type as a number to tantivy. Therefore, I use text to store the id field. For keeping the order, I would like to sort by the id field with text type.

[fulmicoton]

I suggest we just remove the ability to sort the index by something. It has brought more bugs, confusion than any other feature.

[gy0801151351]

Because tantivy has its own docId field, and tantivy's interface does not provide a save method (override by query). This leads to the confusion of query results in concurrent testing (single-threaded modification + multithreaded query). So, I think index sort is necessary.

[fulmicoton]

I don't understand your sentence.

[PSeitz]

For keeping the order, I would like to sort by the id field with text type.

Why do you want to keep order in the tantivy index?

I suggest we just remove the ability to sort the index by something. It has brought more bugs, confusion than any other feature.

I also don't think the maintenance cost of it justifies potential gains currently. In practice there's little to no benefit, but some confusion about it. Range queries may be accelerated by using binary search instead of a full scan, but we don't do that currently. Compression may be improved, we don't have much data about that though.

Tantivy users may have custom queries on top of sorting. Hard to tell if and how they use it, but only performance should be affected when removing it.

If we decide to remove it I would add a deprecation warning in the upcoming release.