Programs

• a new drg subcommand: aiming at reproducing acedrg -c (background and some details in #401)
• gemmi-convert: added option --pysel=SEL

Library

• a lot of new code needed for gemmi drg
• Python selection syntax
• Structure::find_spacegroup(): return nullptr/None for non-crystal structures

Python

• bindings for FlatStructure (some details in #393)
• drop support for Python 3.8, which reached End-Of-Life in October 2025.

Library

• Add calculation of hydrogen bonds according to the venerable DSSP method
  (implementation of the whole DSSP method was considered, attempted and postponed)
• Add reading of MRC map in mode 12 (float16_t)
  (using the half library that's now bundled with gemmi)
• bundled third-party libs: updated fast_float
  and added https://sourceforge.net/projects/half
• Calculating anomalous maps (f") from a model
• gemmi <-> mmdb conversion: handle SEQRES
• to_mmcif: populate more items in _struct_ref_seq.
• to_pdb: to preserve original letter casing of modres.mod_id
• mmcif: read _em_3d_reconstruction.resolution into Structure::resolution
• Add a function populate_structure_from_block() (#383)
• cif.hpp: keywords loop_, global_ and stop_ can be part of unquoted strings
  (stop_ is a keyword, but stop_it is not)
• Added Intensities::merged()
• Added add 6 PYR residues to the internal ResInfo table
• In add_hydrogens_without_positions(): tweaked how occupancies of neighbouring
  atoms are considered
• Tweaked determine_cutoff_radius() in dencalc.hpp

Programs

• blobs: can be called with a map instead of map coefficients
• h: add option --d-fract=FRACT (#385) that adds atoms as H/D mix,
  with the given deuterium fraction.
• sf2map: add option --pow which outputs, for instance, a Pattterson map
• contact: add option --asus that list asymmetric units that are in contact with 1_555,
  not individual contacts.
• sfcalc: Add support for writing anomalous structure factors (FCanom and PHICanom columns)
  to MTZ files alongside regular structure factors.
• fix: gemmi convert --add-tls didn't work for mmCIF input

Python

• Added a small utility fetch.py to fetch files from the PDB:
  can be used as python -m gemmi.fetch 3abc
• Added read_structure_string()
• Added bindings to misc C++ functions
• Added FlatStructure that exposes coordinate data in a single table

Library:

• A breaking change in Op:

  Until now parse_triplet("h,k,l") was equivalent to parse_triplet("x,y,z").
  This is not how it's handled in cctbx and in, for example Pointless.
  In the Pointless documentation of the REINDEX keyword there is such a note:

  Note that the real and reciprocal space operators correspond to mutually transposed matrices, eg "x-y,-y,-z" corresponds to "h,-h-k,-l".

  Gemmi Op was changed to store the notation kind that was used at creation,
  so now parse_triplet("h,k,l") .triplet() gives "h,k,l", not "x,y,z".
  New Op methods have been added: is_hkl(), as_hkl() and as_xyz()
  Apart from this, in case of hkl-Ops parse_triplet and triplet silently transpose the rotation matrix. See #359 for details.
  This change was extensively tested with Mtz::reindex() – against Pointless.
  It could happen that it causes problems in other scenarios.

• DDL2: added alternate (deposition) checks

• Mtz: added C++ Mtz::write_to_buffer() and Python Mtz.write_to_bytes() (used for streaming MTZ files from web servers)
  and C++ size_to_write()
  These new C++ functions are currently undocumented, not sure if they are useful.
  Changed Mtz::write_to_string() to clear the string instead of appending to it. Probably nobody expected the latter.
  Reorganized MTZ reading to avoid fseek() – so we can read a stream or gzipped file, without storing/unpacking it into memory buffer as we did before (i.e. reading huge gzipped MTZ files uses less memory)

• added UnitCell::find_nearest_pbc_images()

• CIF reading: added optional argument int check_level=1 (see docs for details)

• Form factors: support for custom form factors (in the form of sums of 5 Gaussians) – for a Cambridge project on chemistry-dependent form factors.

• to_mmcif: write _atom_site_anisotrop.pdbx_PDB_model_num if there are 2+ models

• to_pdb: split option use_linkr into use_linkr and use link_id

• the built-in residue list is now partly editable (example in docs)

• mmcif: workaround for a problem with reading 7pvv.cif (#369)
  occupancy and B_iso_or_equiv are now optional (because of #375)

• pdb: added option ignore_ter to PdbReadOptions

• started working on secondary structure determination – that's unfinished and unusable yet

• renamed interpolate_grid_of_aligned_model2() to interpolate_grid_around_model()

• added functions interpolate_points() and interpolate_grid_flexible() (#363), needed in Pandda2

Python:

• added python bindings to Ccp4Base::ccp4_header and to some functions related to gemmi::ChemComp

Program

• gemmi-validate: new check for monomer files, and new option --depo to check against PDB's deposition criteria (from DDL2)

Library

• reading mmcif: added reading of TLS information
• writing mmcif: added a few new items
• using Logger also in classes Ddl and CifToMtz (Logger is now used in all library functions that output warnings or messages)
• improved and documented mmCIF validation (with DDL2)
• added read_ccp4_header() for reading only map header when a map is a huge file
• added a few functions related to TLS (will be documented later)
• documented: working with XDS files, normalization of amplitudes (F->E)
• calculating merging statistics (R-merge, R-meas, R-pim, CC1/2) in various ways: Gemmi can calculate R-merge (and other R-*) in 3 different ways that are present in the literature and other programs; for CC1/2, the sigma-tau method is used
• internal refactoring of file reading
• misc bug fixes

Python

• functions Mtz.filtered() and XdsAscii.filtered()
• a number of other additions and a few small changes/fixes
• all cif-reading function consistently read gzipped files (previously, cif.read_file() and gemmi.read_small_structure() didn't)

Program

• gemmi fprime supports ranges of energies
• gemmi merge – added new options, most importantly --stats to print quality metrics
• gemmi convert – option --add-tls to convert "residual" B-factors (from Refmac) to full B-factors
• gemmi mask – solvent masking that takes into account alternative conformers and atom occupancy (experimental)

C++14 (or later) is required to build the library, C++17 (or later) to build Python bindings.
Expect breaking changes, especially in Python bindings.
The lists below are not complete, but should cover most of the changes.

Library

• Added unified logging of warnings/errors from various gemmi functions (class Logger)
• replaced string Model::name with int Model::num
• mmcif: better handling of null auth_comp_id
• fixes for mmJSON
• Removed deprecated functions:
  • UnitCell.fractionalization_matrix and orthogonalization_matrix – use frac.mat and orth.mat
  • count_hydrogen_sites() – use has_hydrogen() or count_atom_sites(gemmi.Selection('[H,D]')
  • Grid::resample_to() – use interpolate_grid()
• unified API of Grid interpolation functions. They now have parameter order that can be 0 (nearest value), 1 (linear interpolation), or 3 (cubic). In C++ there are also functions such as trilinear_interpolation() to ensure no overhead.
• to_pdb: write HET records
• Extended selection syntax with: [metals] and [nonmetals].
• Added function set_is_metal() intended for debatable metalloids
• improved interoperability with MMDB (a CCP4 library)
• MonLib: removed read_cif args
• mtz: fixed writing BATCH records
• hydrogen placement: fixes needed for new files with metals in CCP4 Monomer Library
• pdb: fixed reading TLS S tensor
• Structure metadata: expanded RefinementInfo

Python

• Python bindings migrated from pybind11 to nanobind.
  • Much lower runtime overhead, faster build times, better error diagnostics.
  • Built-in typing stubs.
  • Only Python 3.8+.
  • Sadly, no support for Buffer Protocol. It was replaced with NumPy __array__ methods.
    For NumPy, you can also use .array properties that were available also in the previous releases.
  • No implicit conversions from list to ndarray, and from bytes to string (let me know where it causes problems)
  • gemmi.ValueSigmaAsuData.value_array has now shape (N,2)
• Added pickling support for Structure, Model, Chain, Residue, Atom, cif.Document, cif.Block.
• Added function interpolate_position_array (#323).
• Python extension module is now installed into site-packages/gemmi/ (this change should be invisible to the user)

Program

• gemmi convert --sifts-num is now more customizable
• gemmi sf2map: added option --check (see docs)
• gemmi cif2mtz: add a rule to spec to convert pdbx_F_calc_with_solvent to F-model (+phase)
• gemmi xds2mtz: handles merged files from XSCALE
• gemmi mtz2cif and merge: recognize extension .ahkl as XDS file

This is primarily a bug-fix release. New Python bindings are not included yet.

Enhancements:

• New subcommand gemmi set for changing coordinates, B-factors and occupancies in coordinate files (mmCIF and PDB). Unlike other tools, it replaces numbers while leaving the rest of the file intact. An alternative to CCP4 PDBSET keywords: BFACTOR, OCCUPANCY, SHIFT, NOISE. Note that gemmi convert offers overlapping capabilities. For instance, gemmi convert --apply-symop=x+0.123,y,z shifts the coordinates similarly to gemmi set --shift='9.3 0 0' (the latter takes the shift in Angstroms).

• Improved anisotropic scaling of structure factors. More work is planned in this area.

Fixes:

• fixed reading of mmCIF files without _atom_site.auth_seq_id
• in Topology preparation: fixed a couple of bugs, peptide links are now assumed to be CIS for ω=0±60° (previously, ω=0±30°)
• fixed re-assignment of ATOM/HETATM record types (gemmi convert --assign-records)
• fixed gemmi convert --sifts-num for UniProt sequence numbers >5000

And various minor changes that are hard to describe concisely.

Library:

• SmallStructure: changed how the space group is read and accessed.
  Relying on H-M space group names alone was not always sufficient. The new mechanism uses the list of operations and Hall symbol in preference to the H-M symbol – the order is configurable.
• symmetry triplets: parse decimal fractions (small molecule files may use notation such as x+0.25 instead of x+1/4)
• tabulated space groups: a few more settings: B 1 2 1, B 1 21 1, F 1 m 1, F 1 d 1, F 1 2 1
• X-ray scattering coefficients: changed the default value of IT92::ignore_charge to true (i.e. charges are now ignored by default; before version 0.6.3 they were always ignored)
• cif::Table: added method ensure_loop() that converts tag-value pairs into a loop; might be needed before calling append_row()
• place_hydrogens(): fix for NH3-like configurations
• improved gemmi->mmdb conversion
• Grid: tweaked good_grid_size() to ensure that when creating a grid up to a certain d_min, all reflections up to d_min are in the grid (it matters when no oversampling is applied)
• DensityCalculator: deprecated function set_grid_cell_and_spacegroup(), use grid.setup_from()
• fixed TNT-compatible reciprocal space ASU calculation for non-standard settings
• infer_polymer_end(): complicate the heuristic even more, to detect files that have HETATM incorrectly used for standard residues in a polymer (such files were reported, they are either a result of mutating from non-standard residues, or a buggy program)
• added function assign_het_flags() to re-set ATOM/HETATM flags
• Model: added funtions calculate_b_iso_range() and calculate_b_aniso_range(); the first one can be used to detect if pLDDT is in the range 0-100 (like from AlphaFold) or 0-1 (like from ESMFold)
• writing mmCIF: write _entity_poly_seq.hetero
• added flag Entity::reflects_microhetero that shows if sequences were read from SEQRES (and don't account for point mutations) or from _entity_poly_seq; new function add_microhetero_to_sequences() changes the former to the latter

Program:

• gemmi sfcalc: added a few more options
• gemmi convert: added options --assign-records[=A|H], improved --sifts-num, adding microheterogeneities to _entity_poly_seq when converting from PDB
• gemmi cifdiff: added option -t for basic comparison of values for a single tag

Other:

• minimal WebAssembly port (C++ code compiled with emscripten) of Structure,
  as a proof-of-concept and for reading mmCIF files in UglyMol
• examples/to_rdkit.py: example of conversion of gemmi ChemComp to RDKit Mol

and a number of less important changes

Library:

• gemmi can now be built with zlib-ng, a faster fork of zlib (good for working with large, compressed files)
• experimental: binary serialization of Structure (contained objects, such as Model, Chain or UnitCell, can also be serialized separately)
• finalized handling of 5-character monomer names; uses the tilde-hetnam extension (ABCDE ↔ ~DE) for PDB files
• when atom names in the coordinate file match previous names (_chem_comp_atom.alt_atom_id) from the monomer library (the names in the CCD and therefore also in the ML change occasionally), print better diagnostic; added function MonLib::update_old_atom_names() to update the names in a Structure
• topology: fixed handling of two bonds between the same two residues
• options for handling mmCIF files with incorrect entities (modified add_entity_ids() when called with overwrite=true)
• added function Intensities::prepare_merged_mtz()
• a few bug fixes (for instance, in handling of negative residue numbers in the selection syntax)

Python bindings:

• generating type stubs - see #293
• python: cif.Loop.val() has been replaced with __getitem__/__setitem__
• fixed Mtz.Batch.ints and Mtz.Batch.floats

Program

• subcommand diff has been renamed to cifdiff
• subcommand prep has been renamed to crd
• validate: more options for checking monomer files
• gemmi-grep: added option --extended-regexp
• mtz2cif: added column names Iplus/Iminus (used by ccp4i2) to the default conversion spec

Note: this list is meant to show important changes only.

Library

• completely changed build system for Python module, from setuptools to scikit-build-core
• optimized electron density calculation: single-precision version is now about 2x faster and slightly less exact; some other grid-based calculations also got optimized in the process
• as part of the above optimizations, some of the grid computations require that the model is in the standard orientation (conventional axis directions); in other cases (which are very rare after the remediation of non-standard coordinate frames in the PDB) call standardize_crystal_frame()
• CIF output: more flexible formatting
• mmCIF writing: category _entity_poly is included by default, with pdbx_strand_id and pdbx_seq_one_letter_code
• minor changes in reading mmCIF coordinate files
• cif: added functions Loop::add_columns(), Loop::remove_column(), Column::erase()
• MRC map format: ORIGIN record is ignored (previously, if ORIGIN was non-zero, Ccp4::full_cell() returned false and some map properties were not set)
• new function Grid::symmetrize_avg()
• fixed bug in ReciprocalGrid::prepare_asu_data()
• added function read_pir_or_fasta() for reading sequences (previously it was undocumented and more limited)
• added function pdbx_one_letter_code() which returns a string like AA(MSE)H..., for _entity_poly.pdbx_seq_one_letter_code
• new functions expand_one_letter() and expand_one_letter_sequence() that take ResidueKind.AA/RNA/DNA as argument replaced expand_protein_one_letter*()
• adjusted weights in align_sequence_to_polymer()
• added function assign_best_sequences()
• PDB reading: added Structure::ter_status flag to indicate if TER records were: absent, present, clearly in wrong places
• experimental (not documented yet) new functions: Model::get_cra(), Model::get_parent_of()
• Topo::Bond stores a flag for bonds between different symmetry images
• ChemComp::Atom: store _chem_comp_atom.alt_atom_id as old_id, use it in new function update_old_atom_names()
• riding hydrogens: added H had wrong occupancy in special, rare cases
• added Vec3f – Vec3 with single-precision numbers
• minor API changes: Binner::setup() doesn't return anything, changed argument types of Scaling::scale_data(), align_sequences()

Program

• new tool gemmi-diff that compares categories and tags in two (mm)CIF files
• gemmi-align prints vertical list with option --verbose
• gemmi-residues has new options: -e, -sss, --chains
• gemmi-rmsz: added option --missing to print missing atoms
• gemmi-validate: more options for validating monomer files
• gemmi-h: more options
• gemmi-mtz: prints info about SYMM records

• new: normalization of amplitudes using so-called "Karle" approach, similar as in the CCP4 program ECALC
• added X-ray scattering coefficients for ions (previously, the charge of atom was ignored)
• pdb: reading CONECT records, and an option to also write them
• when reading pdb, if any chain has 2+ TER records, all TER records are ignored
• more configuration options for writing pdb files
• added functions Mtz::expand_to_p1() and Mtz::read_file_gz()
• cif::Block::find_value(tag) now returns also value from the corresponding loop if that loop has only one row
• changes in gemmi-validate related to validation with DDL2
• gemmi-sfcalc: added option --sigma-cutoff
• gemmi sf2map --mapmask: if the unit cells in coordinate file is different than in SF file, use only the latter
• improved transform_to_assembly(), expand_ncs() and rename_chain()
• cif2mtz: Mtz column for pdbx_DELPHWT has now label PHDELWT (#272)
• fixed ensure_asu(): phase-shift (for phases and H-L coefficients) was wrong
• fixed UnitCell::find_nearest_image() for non-crystals with NCS
• fixed DensityCalculator::requested_grid_spacing()
• changes and enhancements in add_chemcomp_to_block(), in solvent masking, in mtz2cif,
  and in several other places
• added python bindings to MtzToCif, cif::Ddl, PdbWriteOptions, changed how options for PDB writing are passed, more bindings for Mtz::Batch