v1.15.0

Fixed

• TUI zebra stripe readable on light terminals. The default theme's
  alternating-row background is a near-black grey that only reads as a subtle
  stripe on a dark terminal; on a light-background terminal (e.g. JupyterLab's
  web terminal) it became a hard black band that swallowed the default-foreground
  text. vv now detects the terminal background (OSC 11 query, then COLORFGBG,
  overridable with VV_BACKGROUND=light|dark) and, on a light terminal, uses a
  subtle light-grey stripe instead. Dark terminals are unchanged; detection
  failure falls back to the previous behaviour.

Added

• --arrow / --feather output — write an Arrow IPC file (Feather v2), the
  zero-copy interchange format for pandas / polars / R arrow. Same column
  projection / --filter / - (stdout) handling as --parquet; streams
  chunk-by-chunk. --compression accepts zstd (default), lz4, or none
  (Arrow IPC body compression). Round-trips through vv and is read by pyarrow.
• --count — print the row count and exit. Instant on formats that carry a
  count in metadata (Parquet, LociSSD); reflects a -r region and, with
  --filter, counts only matching rows. A wc -l for any supported format.
• --describe --json / --describe --ndjson — machine-readable per-column
  statistics (the same count/nulls/min/max/mean/distinct as the text table) as a
  JSON array or one object per line, for reproducible QC in pipelines. Integers
  are exact (not %.6g-rounded), floats round-trippable, strings JSON-escaped;
  numeric columns omit distinct, all-null columns report null min/max/mean.
• Reference-aware pileup — vv x.bam --pileup -f ref.fa. With a reference
  FASTA (needs a .fai), the --pileup output fills the ref column from the
  reference and renders read bases matching it as . (forward) / , (reverse),
  byte-for-byte like samtools mpileup -f (vv applies no BAQ, i.e. -B); deleted
  bases are filled from the reference too. Without -f the previous behaviour is
  unchanged (ref N, literal bases). The current contig is fetched once and
  cached. Verified byte-identical against samtools mpileup -B -f (matches,
  mismatches, deletions, ref-column case). -f without --pileup is a clean
  usage error.

Changed

• LociSSD v4 reads faster. The colblock read_chunk decoded the Start
  column twice whenever End was also present (once for the LENGTH/mask input,
  once as an output column) — the universal case for these interval files. Start
  is now decoded once and the array reused. ~40% less time on both a full
  --tsv scan and a -r region query over a 1 M-row file; output byte-identical.

Fixed

• LociSSD v4 reader hardening (memory safety). The colblock decoders trusted
  on-disk offsets from the (untrusted) file: the DICT and ARENA string codecs did
  assign(blob + o0, o1 - o0) with no o0 ≤ o1 ≤ blob_len check (a non-monotone
  offset underflowed the length to ~4 GiB → out-of-bounds read), n_dict + 1 was
  computed in 32-bit (wrap at UINT32_MAX), a -r region query on a file lacking
  Start/End indexed columns out of range, and zstd_inflate had no output ceiling
  (a tiny chunk could inflate to gigabytes). All now validate and reject with a
  clean error (non-zero exit), never an OOB read or OOM. Crafted-input fixtures
  (tiny.v4.baddict/badarena/nocoord/zbomb.lociss) guard each path.

v1.13.0

A focused follow-up that makes AnnData obs / var actually dumpable as text.

Added

• Full obs/var text dump. vv file.h5ad --tab obs --tsv (and --tab var)
  now exports the complete DataFrame instead of the 1000-row interactive
  preview — previously the dump was silently truncated at 1000 rows. -n N
  limits it; the TUI and plain table view keep the bounded preview. The sparse /
  dense X matrix stays capped (a full text dump of X is intentionally not
  enabled).

Fixed

• High-cardinality categorical obs/var columns now decode to their string
  labels instead of integer codes. The category-dictionary cap was 65536, which
  wrongly showed real columns like CRISPR perturbation guides / targets
  (hundreds of thousands of categories) as raw codes; the default is now
  1,000,000, overridable with VV_CATEGORY_DICT_CAP.
• Boolean obs/var columns (stored as HDF5 enums — e.g. highly_variable,
  mt) render their values (FALSE / TRUE) instead of ?.

v1.12.0

AnnData depth (CSC sparse, uns, labelled X), data-correctness fixes across
the CSV / display / numeric paths, security hardening, and broad RAM/throughput
work on the streaming and TUI paths.

Added

• AnnData uns decoding — a new key/value tab surfaces the unstructured
  annotations: scalars, strings and short arrays show their values, nested dicts
  flatten with dotted keys (pca.variance_ratio), and an encoded sub-object
  (dataframe / categorical) shows its encoding-type. Previously skipped.
• AnnData CSC sparse X preview — csc_matrix matrices now densify to a
  rows × columns preview just like CSR (the compressed axis is columns, not
  rows); they previously showed only a "not implemented" note.
• Labelled AnnData X preview — the sparse / dense X preview names its
  value columns by the var index (gene names) and prepends the obs index
  (cell barcodes) as a row-label column, instead of generic col0 / row numbers.

Fixed

• Leading-zero IDs in CSV/TSV (and spreadsheet / markdown / HTML tables) are
  preserved — a column of values like 007 is read as text instead of being
  inferred as the integer 7. Scientific notation stays numeric.
• Wide and combining characters align. display_width now measures terminal
  columns (CJK / fullwidth / emoji = 2, combining / zero-width = 0) instead of
  counting codepoints, so tables with such data no longer skew; truncation cuts
  on a column boundary.
• Hostile / malformed files are handled gracefully — AnnData sparse shape
  and DataFrame column lengths are validated against the real dataset extents
  (no out-of-bounds read).
• Date / timestamp / decimal columns are read as numeric values, so
  --describe min/max/mean and --heatmap work on them; --describe now
  summarises the whole table rather than the first chunk.
• Genomic / format correctness — Parquet region pruning indexes column
  statistics by Parquet leaf (not Arrow field); pileup honours reference skips
  (CIGAR N) like samtools; mid-stream htslib read errors in pileup / BCF
  region mode are surfaced instead of silently truncating; region mode reports
  exact post-filter row counts; CSV headers of nan / inf / hex are no longer
  mistaken for headerless numeric data.
• TUI / terminal robustness — the terminal is restored on
  SIGINT/SIGTERM/SIGHUP via an async-signal-safe handler; markdown rendering
  strips terminal control bytes; truncate never splits a multibyte codepoint;
  a missing flag argument reports the specific flag instead of "Unknown option".

Performance

• Bounded RAM on streaming sources — DelimitedSource and the other streaming
  readers evict old chunks behind a retention window instead of retaining every
  decoded batch; SQLite computes COUNT(*) lazily (not at open); generic HDF5
  1-D datasets and wide NPZ arrays preview a capped head; delimited preamble
  stripping is buffered. TUI redraws format each visible cell once and reuse the
  LRU chunk cache during sorted search.

Security

• SQLite identifier quoting escapes embedded double quotes, so a table named
  a"b no longer produces malformed / injectable SQL.
• NPZ allocation hint — the zip central-directory uncompressed_size (which
  is attacker-controllable) is no longer passed straight to reserve(), so a
  tiny crafted entry can't force a multi-GB allocation.

Build / CI

• KDE plugins stop including KDEInstallDirs, silencing the ECM directory
  warning in the GUI build.

v1.11.0

Large-AnnData usability, a Debian package, and build/CI maintenance.

Added

• --tab <name> — view a named component tab (AnnData obs / var / X,
  a workbook sheet, …) straight from the CLI, e.g.
  vv cells.h5ad --tab obs -n 20. Case-insensitive; an unknown name lists the
  available tabs. The data components were previously only reachable in the
  interactive TUI.
• Debian package — packaging/debian/build-deb.sh wraps the static
  AlmaLinux 8 binary into a self-contained .deb (Depends: libc6 only), and
  the release workflow now publishes a vv_<version>_<arch>.deb for x86_64 and
  aarch64 alongside the tarballs.

Fixed

• Large AnnData components no longer stall. Opening a multi-GB .h5ad
  (e.g. a 100M-cell atlas plate with a 4.7M-row obs) used to hang for minutes:
  the TUI eagerly materialised every component and read obs/var in full.
  Now each component tab is built lazily (only when viewed), and obs/var/X
  are read as a bounded first-rows preview (footer notes the true size). A
  high-cardinality categorical (e.g. a per-cell barcode with millions of
  categories) is shown as integer codes instead of reading its whole dictionary.
• Arrow deprecation warnings in the Parquet read paths silenced by moving to
  the arrow::Result overloads of ReadRowGroups / GetRecordBatchReader
  (version-guarded so the Arrow 23.0.1 static build still compiles).

Build / CI

• ncurses is fetched from the GNU mirror (ftp.gnu.org) instead of
  invisible-mirror.net, which intermittently WAF-blocked CI runners and broke
  a release.
• GitHub Actions bumped to their Node 24 runtimes ahead of the forced migration.

v1.10.0

A feature and robustness release: a new terminal heatmap view, a major
upgrade to the Qt6 desktop GUI (vvg), and the remaining high-severity
correctness/OOM fixes from the code audit.

Added

• --heatmap — render the numeric columns as a colour heatmap in the
  terminal (rows × numeric-columns, globally normalised, viridis palette).
  --image-mode selects the backend: auto (kitty graphics if supported, else
  half-block), kitty, sixel, halfblock, or ascii. When stdout is not a
  terminal a plain ASCII intensity grid is written instead of raw escape
  sequences.
• Qt6 GUI (vvg) — a real desktop application: menu bar, File▸Open
  (multi-file → tabs), drag-and-drop, recent files, and error dialogs; a
  genomic region/tabix query bar with a pileup toggle (re-opens BAM/VCF/BED
  over a region); a View menu (column show/hide, go-to-row, shortcuts help).
  Filtering, sorting and find now run off the UI thread with a progress bar
  and a Cancel button, so the window stays responsive on large files.

Fixed

• HDF5 / AnnData — a dense X matrix tab is previewed (first 1000 rows ×
  200 columns) instead of densifying the entire matrix into RAM, which could
  OOM/abort the reader on a real dataset.
• Workbooks (XLSX & ODS) — a row wider than the header no longer makes
  Arrow reject the whole sheet; every row is padded to the widest and the
  overflow header columns are named colN.
• ODS — table:number-rows-repeated on a non-empty row now expands to N
  rows (capped) instead of silently dropping the duplicates.
• TUI — sort / filter / stats / search drain a streaming source to EOF
  first, so they cover the whole file instead of only the chunks scrolled
  through so far.
• --heatmap — Inf/NaN cells no longer poison normalization (they're
  treated as gaps); the scan buffer is bounded; non-TTY output is guarded.
• KDE plugins — the Dolphin thumbnailer and KFileMetaData extractor handle
  malformed input gracefully (empty result, no worker abort).

Performance

• GUI — row→chunk mapping is O(log chunks) via a cumulative-offset binary
  search (was O(chunks) per cell); find navigates a precomputed match list in
  O(log matches) and uses an Arrow match_substring prefilter for literals.

Docs

• Documented --heatmap / --image-mode, NumPy .npz / .npy, and .fods
  across --help, the man page, the README, and the bash/fish/zsh completions;
  clarified that the KF6 GUI plugins are optional.

v1.9.1

A correctness- and robustness-focused patch release: it fixes a class of
silently-wrong results, several crash/hang/OOM paths on malformed input found
in a code audit, and the release/CI pipeline. No user-facing feature changes.

Fixed

• Region coordinates — every htslib-backed range query (tabix BED/VCF/GFF,
  BAM --pileup -r, BCF) fed vv's canonical UCSC 0-based half-open coordinates
  to htslib's 1-based-inclusive region parsers, shifting every query one base
  and disagreeing with the Parquet path for the same -r. Coordinates are now
  converted at the htslib boundary so all formats agree.
• Parquet region queries now handle every integer coordinate width
  (UInt32/Int16/…, and dictionary-encoded ints), not just Int32/Int64 — a
  UInt32 Start/End column previously made region queries return nothing.
• SQLite NUMERIC/DATE/DATETIME/BOOLEAN columns are preserved
  verbatim instead of being coerced through double (which turned dates into
  their year and rounded 64-bit integers beyond 2^53).
• BCF keeps the FORMAT field (e.g. GT:AD:DP) in the FORMAT_SAMPLES
  column instead of dropping it.
• Delimited streaming surfaces a mid-file parse error (a malformed row
  beyond the first 16 MiB block) as a non-zero exit instead of silently
  truncating output with status 0.
• Streaming sources (Arrow IPC, FASTA/FASTQ, SQLite) no longer spin or
  silently truncate when a read error occurs past the first batch.
• Empty tabix region (a window overlapping no records) returns an empty
  result with exit 0, matching the Parquet/BCF/BAM paths, instead of erroring.

Security

Memory-safety fixes for malformed/untrusted input (reachable in-process via
the KDE thumbnailer / KFileMetaData extractor):

• HDF5 — array-valued attributes no longer overflow fixed-size buffers in
  H5Aread (a crafted shape/string attribute could smash the stack — verified
  segfault, now rejected).
• NumPy .npz — the declared array shape is validated against the stored
  data (rejecting negative dimensions, overflowing products, and shapes larger
  than the buffer) instead of driving an out-of-bounds read.
• 2bit — the header seqCount is bounded against the file size before
  reserving (a crafted value requested ~170 GB and aborted the process); the
  N-block table skip is computed in 64-bit.

Build / CI

• The KF6 Dolphin/KFileMetaData GUI plugins are now optional in CI so the
  Ubuntu 24.04 job (which lacks the KF6 dev packages) builds the Qt GUI without
  them rather than failing.
• The static-binary release fetch step retries on transfer timeouts
  (--retry-all-errors, connect timeout), fixing the flaky-download failure
  that blocked the 1.9.0 release.

v1.8.2

Fixed

• Static-binary build: the vendored libhdf5 was configured with
  --disable-deprecated-symbols, but main.cpp uses
  H5Dvlen_reclaim to free variable-length strings (its 1.12+
  replacement H5Treclaim doesn't exist on Ubuntu 22.04's libhdf5
  1.10, so the deprecated name is the portable choice). Dropped
  the flag from the AlmaLinux 8 build so the static binary
  compiles. Packaging-only release; no source changes.

v1.7.0

Interactive-viewer focus on top of 1.6.0. Major additions to the
ncurses TUI: mouse clicks, an in-place theme picker with XDG-spec
persistence, a vim-style : command line, and multi-file tabs.

Added

• TUI multi-file tabs — vv a.vcf b.bed c.parquet opens each
  file as a tab. Tab / Shift+Tab cycle. Each tab keeps its own
  scroll position, sort, filter, column-visibility set, search
  anchor, and chunk cache; switching back is instant. Status bar
  shows tab N/M: basename when more than one file is open.
  Non-interactive output modes (--tsv, --parquet, …) still
  process only the first positional — multi-file is TUI-only.
• TUI command line (: key) — vim/less-style typed-command
  prompt at the bottom of the screen. First verbs:
  • :<N> — jump to row N (matches the row-index column;
    out-of-range clamps to the last page). Closes the "scroll
    through millions of rows by hand" pain point.
  • :q / :quit — quit (vim muscle memory).
  • :theme NAME — text-driven theme switch (alt to the T
    overlay), with the same name set as --theme. Persisted to
    ~/.config/vv/config via the standard XDG path.
    Errors stay in the input bar so the line can be edited; Esc
    cancels. Extensible — future verbs (:reload, :w FILE, …)
    slot in without new key bindings.
• TUI theme picker (T key) — overlay listing every built-in
  theme; [*] marks the currently-active one, j/k move the
  cursor, Enter applies the choice. The new theme takes effect
  immediately (ncurses color pairs re-initialised in place) and
  the choice is persisted to ~/.config/vv/config
  ($XDG_CONFIG_HOME/vv/config if set) so future runs start with
  the same theme. Esc closes the overlay without changing.
• XDG-spec user config — vv now reads
  $XDG_CONFIG_HOME/vv/config (default ~/.config/vv/config)
  at startup. Format is plain INI-style key = value; lines
  starting with # are comments. Today only theme is read;
  the format is extensible — future preferences (default
  --threads, --decode-threads, etc.) slot in without breaking
  existing files. Writes are atomic (.tmp + rename) and preserve
  any comments / other keys in place. CLI flags always win over
  the config file.
• TUI mouse clicks — three additive bindings on top of the
  existing wheel-scroll: click a column header to sort by that
  column (toggle asc/desc on repeat; updates the active column
  for S / y); click a data row to scroll it to the top;
  double-click a data row to open the detail pane (same as
  Enter). Shift+drag continues to work as the universal escape
  hatch for native terminal text selection.

v1.6.0

A broad feature release on top of 1.5.0. Highlights: bigBed / bigWig
and 2bit support via a vendored libBigWig and a hand-rolled 2bit
reader; full TUI exploration suite (column stats, sort, show/hide,
live filter, copy cell); generic Parquet range queries with column
auto-detection; markdown output and LociSSD validation; coordinate-
convention selector with UCSC and NCBI labels; color themes; LTO +
ccache static builds; and ARM64 release artifacts.

Changed

• Smarter inline truncation for list / map cells. The
  truncator now walks every top-level comma and picks the largest
  leading-element prefix that fits the column. Previous behavior
  always dropped to [first, …] after the first comma, even when
  [first, second] would have fit. Most visible on VCF INFO
  expansions, BAM tags, and other multi-element list cells common
  in real data. Falls back to char-truncation only when even one
  element + ellipsis doesn't fit.

Added

• TUI copy cell (y key) — copies the top-left visible cell to
  the system clipboard via OSC52. Works through SSH and tmux 3.3+
  with no xclip / pbcopy helper required (modern terminal
  emulators intercept the escape directly). Status bar flashes
  copied: <preview> until the next key; auto-clears.

• UCSC 2bit (.2bit) — sequence container used for genome
  references (hg38.2bit, mm10.2bit). vv exposes the sequence index
  rather than decoded bases (chromosome-scale strings are too large
  to materialise): name, length, n_blocks (unknown-base runs),
  mask_blocks (soft-masked-region runs). Endianness is detected
  from the signature; long-offset 64-bit 2bit is rejected with a
  clear message (use twoBitToFa for that variant).

• --theme <name> — color palette selector. Built-in themes:
  default, dark, light, solarized-dark, solarized-light
  (solarized accepted as a synonym for solarized-dark). Affects
  both the non-interactive ASCII table (ANSI escapes) and the
  ncurses TUI. Each theme provides a 256-color palette and a
  16-color fallback for older terminals; the light theme also
  suppresses zebra stripes so the table reads cleanly on a bright
  background. Unknown theme names produce a clear error listing the
  available choices.

• TUI live filter (& key) — long-requested. Opens a
  &<expression> input bar; expression grammar matches the
  --filter CLI flag. Rows not matching the predicate are hidden
  from the display; status bar shows filter:<expr> N/M (rows
  kept / total). Composes with sort (s): one full-file pass
  rebuilds the combined view via the existing source_row(display)
  indirection. Empty input or Esc (in normal mode) clears.

  Internals: a new rebuild_display_order() orchestrates both
  sort and filter into the single sort_order_ permutation. When
  both are active, rows that pass the filter are collected with
  their sort keys in one pass and sorted at the end.

• --decode-threads N — separately size Arrow's CPU thread pool
  (Parquet column decode and CSV / TSV parallel parsing) without
  touching htslib's thread count. Defaults to --threads; useful
  when cold reads bottleneck on column decompression. Bounded at
  2 × hardware_concurrency().

Changed

• Release pipeline: now builds both x86_64 and aarch64
  (ARM64) static Linux binaries on every tag. The release workflow
  matrices over ubuntu-latest and the new GitHub-hosted
  ubuntu-22.04-arm runner — native builds on each architecture,
  no QEMU emulation. Tarballs are named
  vv-<ver>-linux-<arch>.tar.gz; a merged SHA256SUMS covers both.
• AlmaLinux 8 Docker build: dropped the -march=x86-64
  baseline flag from the build-app CMake invocation so the same
  Dockerfile builds on aarch64. (-march=x86-64 is the default
  target on x86_64 gcc anyway, so the change is a no-op there.)
• AlmaLinux 8 Docker build: rebuild every static dep + vv itself
  with -flto=auto. gcc-ar / gcc-ranlib preserve LTO IR in archives
  so the final link can drop unused code across library boundaries
  (not just within each .a). Significant text-segment shrink vs
  the previous non-LTO 15 MB static binary.
• AlmaLinux 8 Docker build: install ccache (via EPEL) and put
  its compiler-symlink dir ahead of gcc-toolset-12 on PATH. A
  BuildKit cache mount (id=vv-ccache) persists the compile cache
  across docker builds. Requires Docker BuildKit + buildx.

Added

• TUI interactive-exploration batch — three new keybindings turn
  the ncurses viewer into a real exploration tool:
  • S opens a per-column stats popup (count, nulls, min, max,
    mean, distinct count) for the leftmost visible column. Full-file
    scan with progress in the status line.
  • s sorts by the leftmost visible column. Repeated s on the
    same column toggles ascending / descending; s on a different
    column re-sorts ascending; u clears the sort. Numeric types
    sort by raw value, others lexicographically; nulls last. Search
    and the detail pane follow the sorted display order via a single
    source_row(display) indirection that routes all cache lookups
    through the sort permutation when active.
  • c opens a column show/hide picker overlay; Space toggles
    the column under the cursor, hidden columns drop from the
    layout, and the status bar shows a hidden:N indicator.
• --tail N — show the last N rows instead of the first N.
  Mirrors -n but operates on the tail. Reuses the
  MemoryTableSource adapter: every chunk is read (through any
  active --filter) and the last N rows are sliced off, so the
  result renders identically through every view / export mode
  (table, TUI, TSV, CSV, JSON, Markdown, Parquet).
• --coords UCSC|NCBI — pick the coordinate convention for
  -r. UCSC (default; aliases 0-based, bed) means 0-based
  half-open as in BED, bigBed, bigWig, BAM and LociSSD — the
  convention introduced by Jim Kent's UCSC Genome Browser source
  tree in 2000. NCBI (aliases GenBank, 1-based, tabix)
  means 1-based inclusive as in GenBank (1982), VCF, GFF, and the
  samtools / bcftools / tabix command lines. vv -r chr1:101-200 --coords NCBI file matches the same rows as the UCSC default
  vv -r chr1:100-200 file. Conversion happens once in
  apply_region_modifiers; downstream sources see normalized
  UCSC. --regions-file is always BED (UCSC) per the spec.
• --parquet - — write Parquet to stdout. Parquet's
  footer-at-end requires seekable writes, so the data is spooled
  to an mkstemps temp file under /tmp, then streamed to stdout
  and unlinked. Bit-identical to writing to a file path. Enables
  vv ... --parquet - | tool pipelines.
• Generic Parquet range queries — -r chr1:1000-2000 file.parquet
  now works on any Parquet file with chrom/start/end columns, not just
  LociSSD. Chromosome / Start / End column names are auto-detected
  (common variants: Chromosome/Chrom/Chr, Start/POS/
  chromStart, End/Stop/chromEnd). When auto-detection isn't
  clear (or your columns have unusual names) use
  --region-cols Chr,Start,End. Row-group pruning uses Parquet
  ByteArray statistics on chrom plus Int statistics on Start/End;
  per-row filtering inside surviving row groups is correct for
  unsorted files too, just less efficient. Dictionary-encoded chrom
  columns are handled transparently.
• --validate — LociSSD invariants checker. Verifies the
  lociSSD_manifest footer parses, the required columns
  (Chromosome / Start / End / MaxEndSoFar) have the right types,
  the manifest's per-chromosome row ranges cover every row
  contiguously, rows are sorted by (Start, End) within each
  chromosome, MaxEndSoFar[i] matches max(End[chrom_first..i]),
  and every row's Chromosome label agrees with the manifest's
  window. Prints PASS / FAIL per check, caps repeat violations
  at 5 lines per category, exits non-zero on any failure.
• --md / --markdown — write a GitHub-flavored markdown table
  to stdout. Useful for pasting query results into issues, READMEs,
  and reports. Cells escape | as \| and turn embedded newlines
  into <br>. Honours --select, --filter, -n, --no-header,
  -r, --sample. Numbers keep their _ thousands separators
  so the rendered table stays readable.
• bigBed / bigWig support via a vendored copy of
  libBigWig (0.4.8, MIT-licensed,
  compiled with -DNOCURL so only zlib is required). Source lives
  under vendored/libBigWig/; CMake compiles it as a private static
  library (libbigwig.a) and links it into vv. The static AlmaLinux 8
  Docker build picks it up automatically — no new external dependency
  for any distribution channel.
• bigBed: the embedded autoSql definition is parsed into typed
  Arrow columns (signalValue, pValue, qValue, …). A small
  ~150-LOC hand-rolled parser handles primitives, char[N] fixed-
  width strings, enum{…} / set{…} (as strings), and arrays
  (int[N] / int[fieldName] → Arrow list<int>).
• bigWig: rendered as four columns (chrom, start, end, value).
• Range queries (-r chr1:78-99) use libBigWig's overlap APIs
  (bbGetOverlappingEntries / bwGetOverlappingIntervals); no
  external index needed.
• User manual at docs/USAGE.md, with worked examples for every
  flag, every output mode, the LociSSD / tabix / BCF range-query
  flows, and a "by-workflow" cookbook section. docs/build_docs.sh
  renders it to self-contained HTML and PDF via pandoc (with a
  texlive → chromium-via-HTML fallback).
• BCF range queries — -r chr1:100-200 file.bcf now works, using
  bcf_itr_querys over an existing .csi / .tbi index. Missing
  index produces a clear hint pointing at bcftools index.
• --regions-file <BED> — read additional windows from a BED file's
  first three columns. Combines with explicit -r; both a...

v1.1.0

What's new

• --tsv / --csv / --delimiter <sep> — convert Parquet to delimited text, streamed row-group by row-group for memory efficiency
• --no-header — omit the header row in delimited output
• --color=auto/always/never — rich ANSI colors in the terminal, plain text when piped
• RFC 4180 compliant CSV quoting
• Dictionary-encoded columns now show decoded values