Added

• Fetching, caching and display of electron density and cryo-EM maps #1134
• Checksum verification (MD5, SHA-1, SHA-256) in FileDownloadUtils, replacing the previous
  stub. Downloads from files.wwpdb.org and files.rcsb.org are checksummed from their ETag
  without a second request #1133
• HttpStatusException, so callers can tell "not there" from a transport failure #1133
• FileDownloadUtils.downloadFileWithValidation(), which downloads and validates over a single
  connection #1133
• LocalPDBDirectory.getMiddleHash(String), computing the two-character directory from the
  right so that it is correct for both short and extended PDB IDs #1133
• Support for the ECOD distribution format introduced at v294.1 #1141 #1139
• StructureInterfaceList.clusterInterfaces(), which clusters interfaces starting from any
  given clusters and with a custom identifier pair for matching (e.g. ENTITY_ID_PAIR)

Performance

• Contact calculation is about 1.5x faster: squared distances compared against a squared cutoff,
  primitive arrays in GridCell instead of boxed lists, and a pre-sized AtomContactSet #1147
• ASA calculation is about 1.2-1.3x faster, by the same replacement of objects with flat
  primitive arrays in the hot loops #1148
• Interface clustering in StructureInterfaceList.getClusters() uses a greedy leader algorithm
  instead of single linkage, needing at most n(n-1)/2 contact overlap scores and usually far fewer.
  For 1gav (690 interfaces) it takes 45 ms instead of 8.3 s
• EcodInstallation.getVersion() reads the file header instead of parsing every domain. The
  current release is 653 MB and holds nearly three million records #1141

Fixed

• CATH downloads use https and check the response before caching it. The previous http URL began
  redirecting, and the redirect body was cached as classification data #1133 #1138
• CATH parsing no longer throws ArrayIndexOutOfBoundsException on blank or truncated lines, and
  cached files are validated before being used #1145
• Chemical component downloads reject redirects and error responses rather than caching them #1133
• Redirects that HttpURLConnection does not follow by itself - 307, 308, and any that change
  http to https - are now followed #1151 #1149
• FileDownloadUtils: the hash argument was ignored by one overload; downloads could be
  silently truncated; temporary files leaked on failure; validateFile threw on a file with no
  parent directory and on an empty .size sidecar #1133
• mmCIF writer emits the entry identifier as a data item, so getPdbId() survives a
  write-then-read round trip #1144 #1143
• StructureInterfaceList.removeInterfacesBelowArea() clears the cached clusters, which kept
  the removed interfaces
• The ECOD read lock is left balanced when a download fails, so the original error is no longer
  replaced by IllegalMonitorStateException #1150
• FileParsingParameters.setParseCAOnly(true) keeps only C-alpha atoms when reading mmCIF and
  BinaryCIF, as it does for PDB files. Since the unified CIF parser it had dropped only non-CA
  carbons, keeping every N, O and S atom. Groups and chains that hold no C-alpha (waters,
  ligands, nucleotides) are no longer created at all in this mode
• Resolution parsing warns only when the values actually differ

Changed

• createValidationFiles() now defaults to ETagPolicy.USE_IF_HEX_DIGEST, so existing callers
  begin recording checksums where the server offers one #1133
• StructureInterfaceList.getClusters(): interfaces with most residue contacts are taken first as
  cluster representatives, and StructureInterfaceCluster.getAverageScore() is now the average
  score of the members merged into the representative rather than over all pairs of members
• Integration tests run nightly rather than on every pull request #1137 #1135
• Tests migrated to JUnit 5 #1125 #1126 #1038
• Library upgrades #1130 #1132

Removed

• The junit-addons test dependency, no longer needed after the JUnit 5 migration #1126

Fixed

• Fix NPE in Structure.toMMCIF() for some PDB entries (e.g. 2G10)
• Maven plugin duplication in main pom.xml
• Fixes for SonarQube S1155
• Some library upgrades

Fixed

• Edge case in quaternary symmetry calculation #1120

Fixed

• Don't use label_seq_id in mmCIF output for non-polymers #1116

Fixed

• mmCIF parsing: entity type should be case insensitive #1109
• Upgraded to latest version of ciftools-java, fixes mmCIF parsing performance issues in some edge cases rcsb/ciftools-java#13
• Added safeguards for Structure.get*Chain*() methods #1111

Fixed

• More lenient mmCIF parsing to be able to parse PDB-IHM entries, some more edge cases #1108

Fixed

• More lenient mmCIF parsing to be able to parse PDB-IHM entries #1106
• Now AsaCalculator main constructor will use first model only instead of all #1107

Added

• New constructor for AsaCalculator to provide Structure and model #1107

Fixed

• More lenient mmCIF parsing, now can read files produced by Gemmi #1103
• Javadocs: all javadoc lint errors fixed
• Plugin updates and fixed issues in maven release process

Fixed

• Bug in symmetry detection #1101 and rcsb/symmetry#118

Fixed

• Brought back a command line argument for protein comparison tool #1096
• Upgraded the log facade framework to slf4j2 #1094
• Code smell fixes (Sonar issues S2293, S1319) #1095 #1091
• Various small fixes

Removed

• Removed capability of automatic download/caching of PDB archive files in MMTF format, following deprecation
  by RCSB PDB. Code to encode/decode in MMTF format is still in place #1099