Update to Clang tidy 20

[Taepper]

Summary

PR Checklist

• All necessary documentation has been adapted or there is an issue to do so.
• The implemented feature is covered by an appropriate test.

[github-actions]

This is a preview of the changelog of the next release:

0.3.0 (2024-08-06)

⚠ BREAKING CHANGES

• old database_config files might be invalid if they contained insertion columns. Also, we are more prohibitive for ndjson input files, which now MUST contain nucleotide/amino acid insertions for all respective sequences. The insertions action and filter do no longer require a column field.
• return data as NDJSON instead of JSON

Features

• add more tests, make less flaky and viable with large dataset (7772ae3)
• add number of partitions to /info (1e780b5)
• add serializationVersion to SILO output for smoother transitioning to new formats (d3badb6)
• added amino acid insertion search, added many test cases and fixed various bugs (d1e4b2b)
• allow default preprocessing config along with user defined preprocessing config (ee9f20e)
• allow null for sequenceName in insertion contains queries (6dbe251)
• allow reading fasta files with missing segments and genes #220 (8ea9893)
• allow reading segments and genes that are null from ndjson file #220 (d0a3a7e)
• also get Runtime Config options from environment variables (33bdd65)
• also log to stdout (54b8a47)
• also return mutation destructed that does not need to be reparsed (a93abbf)
• Alternative templating of symbol classes (6b61985)
• Better test coverage for SymbolEquals filter (42c685c)
• boolean columns, resolves Boolean columns #384: const declaration (685db9f)
• boolean columns: actions/tuple: update assignTupleField() (b020109)
• boolean columns: add and use JsonValueType (2ad1268)
• boolean columns: add bool to JsonValueType, update tuple (3c990e2)
• boolean columns: add expression_type "BooleanEquals" (b82ec69)
• boolean columns: add filter_expressions/bool_equals (ff2a138)
• boolean columns: add optional_bool (a2e47f8)
• boolean columns: add storage/column/bool_column (5eb3430)
• boolean columns: column_group: update ColumnPartitionGroup (d7adecf)
• boolean columns: column_group.h: add {ColumnPartitionGroup,ColumnGroup}::bool_columns fields (af44f1a)
• boolean columns: database (c555a68)
• boolean columns: database_config: add "bool" case to DatabaseConfigReader::readConfig() (12fc7b8)
• boolean columns: database_config: add "boolean" case to de/serialisation (e6d5363)
• boolean columns: database_config: add BOOL to ValueType (0b47b82)
• boolean columns: database_config: update DatabaseMetadata::getColumnType() (30febdd)
• boolean columns: database_partition (0aef8f0)
• boolean columns: optional_bool: add == (f0aa3e8)
• boolean columns: selection (e52f2dc)
• build metadata in parallel to sequences. Do not create unaligned sequence tables in preprocessing, rather hive-partition them directly to disk. Better (debug-)logging (c1cdfeb)
• bulk Tuple allocations now possible (902ec04)
• bump serialization version to 2 (e897203)
• bump serialization version to 3 (7584214)
• change base image to ubuntu (64dee0d)
• clearer Operator::negate and Expression::toString, logical Equivalents for debug printing/logging for the Leaf Operators IndexScan and BitmapSelection (026b639)
• consistent behavior of configs when starting SILO with both --preprocessing and --api (847ec7e)
• declutter README.md from linting option, which is now disabled by default and enforced in the CI for the Linter (9220435)
• details no longer shows insertions (#354) (473cd98)
• display database info after loading new database state (0249416)
• display preprocessing duration in logs in human-readable format (not in microseconds) #296 (a2499af)
• do not enforce building with clang-tidy by default. Linter will still be enforced (7134e45)
• faster builds by copying @corneliusroemer image caching for our dependency images, which rarely change (#374) (7867bc7)
• generalize mutations action to have consistent behavior for different symbols (9834aea)
• generalizing symbol and mutation filters. Clear handling of ambiguous symbols (aa9ad4d)
• have no default sequence by default, implement default amino acid sequence from config (2edf924), closes #454
• have structured and destructured insertion in insertions response (0a7e46a)
• hide intermediate results of the preprocessing - don't put it in the output (44327b0)
• implement basic request id to trace requests #303 (4defb59)
• implement data updates at runtime. More resilient to superfluous or missing directory separators (dc5dfaa)
• improve loadDB speeds (2b7cd7d)
• improve validation error message of some actions on orderByFields (a0da5b5)
• insertion action targets all insertion columns by default (6b70241)
• insertion columns for amino acids and multiple sequence names (3cc8fee)
• insertions action (e067062)
• insertions contains action now targets all columns if the column name is missing (32a6951)
• insertions no longer in metadata or databaseConfig, instead expected for all aligned sequences #372 (1f3680c)
• introduce new storage type for Sequence Positions, where the most numerous symbol is deleted (6e15204)
• introduce storage of unaligned sequences from either ndjson file or fasta file and make them queryable via the Fasta action (44df849)
• load table lazily. Unaligned Sequences do not need to load the table (c2a8439)
• logging improvements (4c12a88)
• make SILO Docker image by default read data from /data (e83b910)
• make threads and max queued http connections available through optional parameter (3ecde68)
• migration to duckdb 0.10.1 (c1426ef)
• mine data version at beginning of preprocessing (362fe0f)
• multiple performance improvements for details endpoint (28f41d0)
• optimize bitmaps before finishing partition (5b06d58)
• reintroduce randomize for all query actions (166045c)
• reserve space in columns when bulk inserting rows (e3c9620)
• return data as NDJSON instead of JSON (c236ba4), closes #126
• return only aliased pango lineages (abf0844)
• reveal some more details when reading YAML fails (1f8d9db)
• save database state into folder with name <data version> (41923eb)
• set the default sequence when there is only a single sequence in the reference genomes (0ae4022), closes #454
• statically disable deleted symbol optimisations because of performance penalty (4e522f4)
• stick to the default of having config value keys in camel case (a1cae40)
• streaming: fasta_aligned: process in sub-partition batches (e87d927), closes #112
• streaming: fasta: process as a batch per partition (66bfdff), closes #112
• streaming: fasta: process in sub-partition batches (12f873f), closes #112
• streaming: fasta: remove limit on total row count (df185cc), closes #112
• streaming: implement materialization cut-off (fd1a77c), closes #112
• support SAM files as sequence input and allow partial sequence input with an offset (5be9a9f)
• template class for sequence store (cef4d48)
• templatized Symbol classes (6b9d734)
• test set with amino acid insertions (de2c4f8)
• throw an error when there is not initialized database loaded yet #295 (b17f72a)
• throw error in preprocessing when default sequence is not in reference genomes (287bb8b), closes #454
• tidying up CLI and file configuration for runtime config. Added option for specifying the port (c3a88a0)
• Unit tests for Tuple (4fc06e8)
• use 'pragma once' as include guards instead of 'ifndef...' (bc49aa5)

Bug Fixes

• adapt randomize query results to target architecture. x86 and ARM have possibly different std::hash results (#355) (600000d)
• add bash dependency which is required by conan build of pkgconf and is not installed on alpine by default (1b3f51c)
• add missing file for test (262bedc)
• add missing sequenceName field to mutation action "orderBy" (06c8c86)
• add workaround so insertions are read correctly (8a2bfa8)
• allow all sequence-names by escaping them properly in all SQL statements (901fc7e)
• allow insertions that start at 0 (#449) (f427137)
• allow sql keywords for metadata field names #259 (6fbeee5)
• also consider 'missing' symbols in the mutation action. Bugfix where Position invariant was broken because of 'missing' symbols (fab72a6)
• alternative non-exhaustive three mer index (467086f)
• always build dependency image for amd64 platform (4f73cb0)
• be able to start without genes, without nucleotide sequences or with neither (e878ed5)
• bug when filtering for indexedStringColumns which are not present in some partitions (62d4e08)
• bug where sequence reconstruction is false when the flipped bitmap is different from the reference sequence symbol (edac58c)
• change random ordering for gcc hashing (78055ff)
• change test to reflect new optimisations (cb63010)
• do not use std::filesystem::path::relative_path() to also support absolute paths (b9ff422)
• don't abort when reading table in chunks (26b1558)
• empty input without partitioning (5fa3c92)
• empty ndjson input files more robust (#473) (9d4232b)
• endToEndResults (c807a33)
• erroneous file created during unit tests should not leak (1b764af)
• error when the Mutations action looked for sequences but the filter was empty (91e5d52)
• floatEquals and floatBetween with null values (47b436e)
• hide nucleotide sequence for default sequence (584715c)
• insertion column, remove reference (d940c52)
• insertions being added at wrong index for large files (#472) (e056ed9)
• linking error on linux (ad9076c)
• linter (54d3c34)
• Make C++ flags in CMake compatible for MacOS (38cae69)
• metadata info test accessing getMetadataFields output no longer directly but over the address of a const (0dcdee1)
• missing include (0773999)
• more efficient ndjson emptiness check (#481) (344ec7b)
• new linter errors (ea6934c)
• nodiscard (silence warnings) (69421c1), closes #390
• non default unaligned nucleotide sequence prefix (93b4829)
• nucleotide symbol equals with dot (6ad623e)
• pango lineage filter with null values (b7238a8)
• put compressors into sql function to avoid static variables (c1a11c8)
• query_handler: check for I/O errors when writing to the HTTP stream (de2b475), closes #112
• quoting {} in "x.{}" SQL struct accesses, as a string starting with a number leads to parser errors (#409) (f3ba6db)
• random (but deterministic for a version) result can depend on internal state, which was changed with duckdb update (9d9351b)
• recursive file reading for nodejs<20 (9ac0ffe)
• reduce function calls and parallelism when inserting rows into storage backend (#509) (93ba858)
• remove caching for linter. Docker image to large on github actions. (8eaeee6)
• remove incorrect compile flags (b92ee4e)
• return instead of continue (03a500a)
• revert duckdb migration due to it being unable to build the new version on the GitHub runner (2a62c54)
• revert test numbers to pre-optimization (99b600a)
• single partition build fix (a8af1c9)
• specify namespace fmt in calls to format_to (#353) (62ffa3b)
• specifying apk versions (2c2354e)
• start with empty files without throwing an error (d407b92)
• streaming: report an error when sorting requested while streaming (313d6bb)
• streaming: report an error when the unimplemented offset or limit matters (3be9c06), closes #511
• test with deterministic results, remove 2 unused variables (96a424b)
• trigger release due to refactor commit-5e3041e5 (#494) (#496) (ecd4ba0)
• unit tests and mock fixtures (db506a1)
• update cmake version on ubuntu (227a2dd)
• Upper and lower bound should be inclusive in DateBetween filter (53c6c05)
• Wrong compare function used in multi-threaded case, which displayed wrong tuples in the details endpoint when a limit was used (650bf36)

[pflanze]

LGTM.

[pflanze]

PS. As discussed, the plan is to go with this as long as clang-format can be kept at version 17. If that doesn't work anymore, either revert the change to clang-tidy 20, or figure out how to configure clang-format ~exhaustively so that changing clang-format version does not yield formatting changes.

Discussion of the latter: https://stackoverflow.com/questions/59395507/how-to-maintain-a-clang-format-file-for-different-clang-format-versions
Problematic point re clangd: clangd/clangd#1999

Could potentially also generate the .clang-format file from our explicit options and public metadata (is https://clang.llvm.org/docs/ClangFormatStyleOptions.html generated from such metadata?) about the options accepted by the newest version of clang-format; but hopefully someone has already written such a generator, research this.