This is a substantial release representing more than 575 merged pull requests. It modernizes the C++ codebase, dramatically expands test coverage, fixes a long list of crashes and correctness bugs, and adds several new options, directives, and commands.

• Fix many crashes in the Python bindings and embedded interpreter: segfaults in journal.query() with --pivot, --account, or period flags such as -M (issues #1193, #1858); use-after-free in Python journal queries (issue #2152); segfaults in close_journal_files() when Python holds commodity references (issues #977, #978); a dangling session pointer when ledger is imported as a Python module (#2163); the segfault when ledger python was reinitialized in interactive mode; and a stack-smash in the Python optional converter

• Fix null pointer dereferences in get_code(), post_pred (issue #1219), apply directive functions (#1221), and magnitude() when a posting has a null amount (#2212)

• Fix dangling context pointers from stack-allocated bind_scope_t and from tag-assertion expressions in the journal

• Fix segfault when a posting iterator is invalidated by a remove, when source_context encounters very large transactions (#1838), in find_price after close_journal_files() left the graph index stale, when the csv --dc format is selected, when ledger starts in a non-existent directory, when --limit filters a commodity inside --budget (#2247), when --gain --subtotal visits multi-commodity accounts (#965), when --sort date runs with bucket accounts (#3128) or against the balance report (#3126), and when an = divider line appears in the journal (#1182)

• Fix assertion failure in truncate() with --payee-width=1, in expr_t::compile for empty expressions (#1899), in alias() when a commodity name already exists, on whitespace-only lines in parse_posts, and from --trace interacting with include directives

• Fix buffer overread in the quantity punctuation-stripping loop and add numeric_cast guards against narrowing overflows

• Fix infinite recursion with --unround and --group-by (issue #518) and the infinite loop produced by a self-referential C (commodity equivalence) directive (#1065)

• Fix --invert to work correctly with composite reports (#1908) and with the equity command (#1645)

This release migrates the codebase to C++17 and replaces several Boost utility components with their standard-library equivalents.

• The build now requires a C++17-capable compiler (GCC 8+, Clang 7+, or MSVC 19.14+). Older toolchains will no longer build

• Boost dependencies are slimmed down. boost::filesystem, boost::any, and boost::variant are replaced with their std:: equivalents, and boost::shared_ptr use is reduced significantly across the core code. find_package(Boost ...) now requires only date_time, iostreams, and regex, with optional python, nowide, and unit_test_framework. The minimum Boost version (1.72.0) is unchanged

• Embedders should note that most public headers now traffic in std::optional, std::any, and std::variant rather than their boost:: counterparts; downstream code that named the Boost types explicitly will need a one-line typedef or search-and-replace

• Adopt C++17 idioms throughout: structured bindings, if-statement initializers, if constexpr, compact nested namespaces, inline static constexpr class constants, [[nodiscard]] and [[fallthrough]] attributes, and std::string_view on parsing hot paths

• Replace raw new/delete with make_shared/make_unique, BOOST_FOREACH with range-for, and throw() with noexcept

• Several memory leaks and a command-injection vector in getquote (when commodity symbols contained shell metacharacters) were closed during the security-hardening pass

• Implement automatic FIFO/LIFO commodity lot matching (issue #164): selling a commodity without an explicit {cost} annotation can now pair the sale with open lots automatically and compute gain/loss from the matched basis

• Add --lots-fifo and --lots-lifo options (which imply --lot-prices and --lot-dates) to display lots in chronological order

• Fix many issues with lot annotations: preserve lot annotations through commodity reduce/unreduce (#958), in balance assertions with annotated amounts, during balance assignments, and in print output; round-trip the fixated cost marker = in print (#1794); preserve {{total}} lot-price notation; preserve user-specified lot dates during transaction finalization; preserve lot price precision when total cost notation is used; derive cost basis from lot price annotations (#630); allow lot_price, lot_date, and lot_tag to accept zero arguments; and use effective dates for lot annotations under --effective

• Fix --lots with commodity conversion (#512), match lots across computed and typed annotation precision (#524), preserve --basis precision when the commodity has none (#3187), and make --price report the @ transaction price rather than the lot basis (#1806)

• Fix transaction balancing for fixated lot prices with two commodities, fixated costs that produce different amounts (#1122), and silently-accepted imbalanced transactions with lot-price annotations

• Fix unrealized gains/losses when multiple prices are recorded on the same day (#1821); add --plopen for unrealized P&L on open positions; add --gain-since for period-relative gain calculation; account for quantity precision when rounding gain/loss; fix the rounding error in --price with @@ total cost notation; warn when a getquote script fails or returns no usable price (#2335); and trigger a price download even when no prior price exists (#652, #655)

• Set commodity precision from price records and let later transactions widen, but not narrow, the recorded precision (#1190, #2481)

• Fix -V failing when a lot-annotated cost marks the cost commodity as PRIMARY (#1217); fix -V -M dropping revaluations in periods without transactions (#2131); fix -H with -X when a posting predates all price history (#1738); fix duplicate commodity quote downloads (#996); make --end exclusive for market-value price lookups (#1762); and walk the larger-unit chain when no direct price is found (#2017)

• Fix --exchange (-X) for quoted commodities (#2321), the multi-commodity balance conversions in get_commodity() (#763) and nail_down() (#717), the C directive interaction with -X (#1066), and crash-when-empty-commodity in getquote (#1869)

• Add --decimal-places to limit display precision (#797), accept an optional precision argument in round() (#628), and add a --round option that rounds each posting before aggregation (#781) plus a --round option that complements --unround

• Allow parenthesized expression prices in P directives (#1101), emit pricedb entries in both directions (#779), filter built-in commodities from pricemap output (#738), report all priced commodities in prices/pricedb (#739), report errors for self-referential price directives, and add --latest to prices/pricedb

• Fix the D directive to lock the commodity display format (#1197), apply D-directive formatting to bare amounts (#761), and allow apostrophe (') as a digit-group separator

• Add batch commodity quote fetching to amortize the per-call cost (#588); add --account-value-expr and --commodity-value-expr options for custom valuations; expose a capital_gain value expression variable

• Filter balance assertions by date rather than file order; skip balance assertions on UUID-duplicate transactions; check account assertions on auto-generated postings; respect --permissive for balance assignments with elided amounts (#2005). Add --check-in-file-order to restore the pre-3.5 file-order accumulation for journals that depend on out-of-order dated entries (#3186).

• Fix balance assertions for accounts with only virtual postings (#1699), with quoted commodities (#904), with per-unit price annotations (#1125), with --decimal-comma (#1167), and with zero-difference values (#1168)

• Fix balance assignments across multiple files (#1831); print balance assignments as = amount rather than amount = amount (#2100); fix balance-assignment alignment in print; honor --effective date ordering in balance assertions (#2071); and use fixated price annotations for multi-commodity balance checking

• Optimize the register command for a 46% speedup on large journals (≈50K postings: 6.0s → 3.3s on the benchmark) by short-circuiting the trivial amount expression, exploiting strip(A+B) = strip(A)+strip(B) to maintain a stripped running total, and early-returning from balance_t::strip_annotations when nothing carries annotations

• Route the accounts, payees, tags, and commodities listing commands through the stats fast path (#3202). These commands only iterate the posting stream to collect distinct names; previously they paid for the running-total and rounding filters used by register and balance, which on a 20K-posting journal made them 6-10× slower than stats. The fast path still honors --limit, --period, --begin/--end, --pivot, --anon, --budget, --forecast-while, and --group-by

• Align multi-commodity balance rows with wide decimals (#1795); fix multi-commodity balance formatting width (#698); make separator lines adapt to amount_width; show the payee when it changes between sorted postings; show zero commodity amounts with --empty in the balance report (extending #2204); preserve virtual posting flags in equity output

• Sort by converted value when -X is active (#1556); sort parent accounts by family total (#723); allow --sort and --sort-xacts together; show the own amount instead of the family total in --flat balance mode

• Enable ANSI color on Windows CMD/PowerShell (#2110); fix justify() to ignore ANSI escape codes when computing width; flush stdout after each REPL command (#991)

• Fix percentage amounts to act as fractions in arithmetic; fix width options to return integers for expression arithmetic; fix the truncated() function to work with amounts in value expressions; expose find_accounts_re() for full regex-matched account lookup

• Make the prices -A (running average) and --deviation options affect the displayed price column (#753); honor --date-format and --datetime-format in prices/pricedb (#1202); make the stats command honor report filters (#742) and use an inclusive day count for single-day journals (#740); exclude future-dated postings from recency statistics (#741); fix bad pricedb info under --anon (#756); fix --anon balance accumulation (#617); add --csv-separator for convert; add --no-group-by to cancel --group-by; fix the misaligned separator dashes in the cleared report (#729) and the bold formatting for uncleared posts under --payee (#865)

• Add the --append-format option (#1740); add --ignore-diacritics for accent-insensitive search; add --pivot-only to replace the account name with the pivot value (#1153); add --period-shift for shifting period boundaries (#954); add --group-by-cumulative for balance sheets accumulating over time

• Rewrite the query parser as a documented recursive-descent parser with a published EBNF grammar in the header

• Fix --empty to respect the query/--limit predicate (#3189); preserve backslash escapes in regex patterns; preserve push-command positional query args across REPL commands (#528); fix the query lexer that included trailing whitespace before operators; handle == in query metadata predicates (#842); honor not in positional queries for print/xact/dump (#1227); evaluate the expr any() keyword in single-argument queries (#1185)

• Expand account aliases in CLI query arguments; fix --pivot with command-line account queries (#1154), with built-in properties (#1048), and with tag-based grouping (#1042); fix --group-by combined with --pivot dropping subsequent groups (#1034); fix --depth interacting with --effective and with --flat; fix --only so it applies after calc_posts and does not alter running totals (#762); compute per-account averages for period reports (#1950); anchor interval averages to the report begin date

• Accept a time component for --begin/--end/--now (#1696): the parser detects datetime structurally and accepts @ as a date/time separator

• Add %m-%d as a built-in short date format (#566); accept relative date expressions (yesterday, tomorrow, tday, yday, tmrw, plus the full period grammar via on <expr>) in the xact command (#647); recognize aux/aux_date/auxdate aliases for the CSV posted column

• Accept duration units (s, m, h, d, w) in --price-exp / --leeway / -Z (#607); document that --price-exp is in hours by default (#1210); apply month rollback for begin dates without an explicit year; fix -b MM/DD to use the most recent past occurrence (#1095); honor LEDGER_INIT_FILE and LEDGER_INIT environment variables (#1189); fix Windows path expansion on Cygwin (#1188); recognize Windows drive-letter paths in the include directive (#1670)

• Resolve date to the latest posting date for accounts; expand bracketed dates in notes during print (#622); fix --input-date-format interaction with xact; fix the --current option being overridden by --end; pin evaluation dates in time-sensitive tests for reproducibility

• Fix -V -M revaluations in periods without transactions (#2131); fix --monthly prices to show the last price per period (#583); restore the year_directive_year when ending an apply year directive (#2413); year directive no longer leaks into xact/entry commands (#707); fix the year directive interaction with price history (#1748); restore year/epoch state after processing included files (#1979)

• Overhaul the forecast feature: forecasts now include the period containing --now rather than only emitting strictly-future periods; mixed-frequency periodic transactions (e.g. monthly plus yearly) are ordered correctly; future-dated from/since and past until clauses on periodic transactions behave correctly; --forecast-while is respected by both balance and register reports; the 5-year safety limit no longer underflows on current-period dates (closes #591, #1044, #1113, #1141, #1148, #1155, #1161, #1576, #1605, plus crash #2043)

• Add payee-rewrite and account-rewrite directives (with matching --payee-rewrite and --account-rewrite CLI flags) that apply regex substitutions to payee and account names at report time without modifying the journal

• Add directives to enable and disable automatic transactions

• Allow $expr in automated-transaction account names in addition to $account; honor effective dates in automated transactions; fix payee-based Unknown account resolution; fix date comparison operators in automated-transaction predicates; include auto-transaction postings in --related output (#993)

• Default to monthly for budget periods without an interval; sort postings by date before budget processing (#590); fix budget-report period initialization with account queries (#636); preserve day-of-month for monthly budget periods with a from clause; honor start dates for weekly periodic transactions

• Apply payee UUID mappings during CSV import; balance timeclock transactions against the bucket account; pre-register built-in metadata tags so --strict no longer warns about them; suppress the spurious --strict warning for $account in automated transactions (#545)

• Strip journal-syntax quotes from commodity names in CSV output (#2008); set the parse context during CSV convert to prevent a segfault under --strict (#2521); fix the CSV mark in description when a code is present; add --csv-separator; honor CSV aux/aux_date/auxdate column aliases for posted dates; fix CSV convert cost-field handling (#1110); apply payee UUID mappings during CSV import; fix CSV convert handling of missing or same-commodity cost fields; fix signed debit amounts in CSV convert; report a clear error for CSV files missing a date column

• Add --time-round to round timelog durations to the nearest unit (#1608)

• Fix the datetime field to include the time portion for timelog postings (#2264); fix payee sub-directive application to timelog entries (#1211); fix the future check-in error in timelog auto-close (#1689); balance timeclock transactions against the bucket account

• Correct --time-colon arithmetic in budget and other computed reports; fix -X to convert time values across scale units (#1177)

• Allow ledger to be loaded as a Python extension module (#513); install the Python module with executable permissions (#531); add --python-home and --python-version to acprep; document how to build the Python bindings; fix the Homebrew build with Python support (#2267)

• Make xacts/posts iterable as Python properties (#682); fix Python xact.posts to return only raw postings (#2453); enable Commodity == string comparison in Python (#759); expose display_amounts on PostCollectorWrapper for commodity exchange (#2158); fix Python FileInfo.filename and FileInfo.modtime access (#1205); fix Session.read_journal() to accept a string path; reset journal state before re-reading from Python; expose price_point_t to the Python bindings

• Fix Python bindings for virtual posting flag detection (#2169) by binding to the correct supports_flags<uint_least16_t> instantiation; fix Python TypeError when accessing boost::optional fields like note; use checked utfcpp functions for Python string conversion; restore terminal settings after ledger python exits (#774)

• Add payee-rewrite and account-rewrite directives for report-time transformation

• Add directives to enable and disable automatic transactions

• Support nested account declarations (#877) and metadata tags on account directives (#1681)

• Return regex capture-group text from the =~ operator (extending #664)

• Major overhaul of the manual: rewrite of the Value Expressions chapter with a full language reference and runnable doctests (#804); document display_amount, display_total, and value_date; expand the value expression reference with previously-undocumented keywords; document the expression engine, time, and query layer (Session 3); document the core data model, journal model, and build infrastructure; document the filter pipeline

• Add real-world examples for tax year grouping with --pivot, recording stock dividends, sales-tax automated transactions, and tag-filtering in the Metadata section (#1117); document account date checks for closing accounts (#1643); document account("name").total for automated transactions; fix the outdated documentation for single-letter value expression variables (#994), the contradictory documentation for --price-db and --download (#1210), and the ansify_if two-argument signature; remove documentation for the unimplemented capture directive

• Switch the default branch from master to main

• Add gpgmepp (the C++ binding for GPGME) as a build dependency when GPG/MIME support is enabled. Builds without -DUSE_GPGME=ON are unaffected

• Drop the legacy AppVeyor pipeline in favor of a GitHub Actions MSYS2/UCRT64 Windows build (#1791); MSYS2/UCRT64 is now the supported Windows toolchain

• Add CPack .deb and .rpm package generators (make package)

• Set up clangd for LSP-based development; add a lint CMake target driving clang-format, clang-tidy, and cppcheck; add lefthook so the same configure/build/ctest/doxygen pipeline runs locally and in CI; show the build type (Debug/Release) in --version output

• Improve PCH configuration and trim system.hh bloat; fix CMAKE_BUILD_TYPE declared as BOOL instead of STRING; fix ICU include path and sysroot clearing for Nix re-runs; build the doc target by default when BUILD_DOCS is ON; update bundled utfcpp to 4.0.9; add explicit CMAKE_OSX_SYSROOT for reproducible macOS builds; add code-coverage tools to the Nix development environment

• Add 419 new regression tests, raising source-code coverage to 89.6%, and lift the CI coverage gate to 90% (#2832, #2728); add regression tests for 153 previously-uncovered open issues (#3011)

• Add a libFuzzer harness for the journal parser, run nightly in CI; add an ASan/UBSan sanitizer build to the CI matrix; add Clang-based performance regression checks; add gcov/lcov code coverage reporting; free disk space and shrink coverage profraw footprint to keep the coverage job within runner limits

• Add a snapshot behavioral testing tool; harden the Python test harness for Python 3 correctness; remove the 4096-byte line length limit in the parser

• Split textual.cc into textual.cc, textual_directives.cc, and textual_xacts.cc along responsibility boundaries, with a shared textual_internal.h

• Rewrite the query parser as a documented recursive-descent parser with structured control flow

• Apply clang-format to the entire C++ codebase; enforce formatting in CI (#2499); reformat several header files for readability

• Refactor a compute_balance_diff helper to eliminate duplication; convert nested namespaces and template specializations to compact C++17 forms; update the fuzzer to use std::filesystem after the boost::filesystem migration

• Fix version number in binary

• Fix crash in optimized builds when exceptions are thrown (issue #693)

• Fix segfault in metadata validation (issue #1800)

• Fix divide by zero (bugs #777 and #2207)

• Fix use-after-free with regex_match()

• Fix bucket transactions ignored with reg --related (bug #2220)

• Restore expr context after calc (bugs #2330 and #2343)

• Fix leap year dates with apply year directive (bug #2468)

• Fix xact command to properly respect last year directive (bug #2413)

• Fix timelog balance assertions with value conversion

• Add support for hash chaining with --hashes option for transaction integrity verification

• Add commodity swaps feature over a base commodity

• Add ==~ operator for regex capture group extraction (issue #664). Without capture groups the operator yields the entire match as a string; with capture groups it yields a sequence whose first element is the entire match followed by every captured group (so a pattern with N groups yields a sequence of length N + 1). Returns false on non-match.

• Add --align-intervals option for reporting

• Add --lisp-date-format option for lisp output (bug #1206)

• Add yday shortcut for "yesterday"

• Allow nested account sub-directives inside an account block so a parent account can declare its children without repeating the full colon-separated path (issue #877)

• Skip files with invalid UTF-8 names in include directive (bug #2421)

• Allow automatic transactions to create accounts with --pedantic (bug #2417)

• Make pricedb output deterministic (bug #1783)

• Update required Python version to 3.10 (minimum 3.9)

• Fix Python 3.12 deprecation warnings

• Add Python bindings for commodity value_t and roundto function

• Remove Python 2 support from tests

• Update to C++17 standard

• Update to Boost 1.86.0 (minimum 1.72.0)

• Update required CMake to 3.16.2

• Add support for Ubuntu 24.04 Noble

• Add build support for readline

• Display parent accounts with one child if they contain posts

• Align amounts even when account names are long

• Fix percentile balance report of time commodity

• Fix divide by zero (#777, #2207)

• Increase string size limit in src/unistring.h assert (#2174)

• Require tzdata for Nix flake build (#2213)

• Fix regression leading to incorrect error about format directives (#2205)

• Add information about compile features to --version

• Fix compiler warnings by minimizing the use of deprecated APIs

• Update flake.nix to match nixpkgs ledger/default.nix

• Remove unused Python server related code

• Various documentation improvements

• Use $PAGER when environment variable is set (#1674)

• Make --depth correctly fold postings to accounts of greater depth into the parent at the specified level (#987)

• When using wild-cards in the include directive, include matched files in sorted order (#1659)

• Ensure absolute path for include (#2075)

• Try to use $XDG_HOME_CONFIG/ledger/ledgerrc or ~/.config/ledger/ledgerrc first

• Improve Python 3 support and drop support for Python 2

• Add support for automatically reading files encrypted with GPG (#1949)

• Add support for a "debit" column in the convert command (#1120)

• Fix parsing of files without end of line (#516)

• Fix incorrect parsing of expressions containing a - without spaces (#2001)

• Fix payee metadata on postings not being validated and payee aliases not being honored (#556, #1892)

• Fix ledger interpreting a posting with 0 difference as a null-posting, which leads to it auto-balancing the posting (#1942)

• Correctly escape all string values in lisp report (#2034)

• Fix a regression where empty commodities were shown (#1969)

• Fix a regression where using multiple commodities in one transaction triggers an assertion (#1998)

• Fix --time-colon for negative time amounts

• Use correct int return type for stream input operations (#2058)

• Use amount_width for balance report

• Remove some UTF-8 code that was having no effect (#2061)

• Fix unrounding for equity

• Fix SIGABRT when python subcommand raises an exception

• Improve XML reports

• Support building on older versions of CMAKE (less than 3.7)

• Fix compilation with Boost 1.76 (#2030)

• Fix Msys2 MinGW build (#1905)

• Fix unicode problems on Windows (#1986)

• Fix the issue that with Boost >= 1.77 include directive cannot find the file to include for stdin (-f -). Also for -f - when include cannot find the file it reports the error with full path now. (#2057, #2092)

• Fix Nix build

• Rename quoted_rfc4180 to quoted_rfc, as numbers used in function names confuses the parser (#2007).

• Numbers are no longer permitted in value expression function names.

• Various documentation improvements

• Fix regression with expression evaluation by reverting commit Correction to the way parens are parsed in query expressions (#1894)

• Fix --invert breakage by reverting commit Change --invert to invert displayed amounts and totals, not amounts (#1895)

• Fix performance regression by reverting commit Compare price annotations using their textual rendering (#1907)

• Fix library path issue (#1885)

• Allow specifying the Python version (#1893)

• Some documentation fixes

• Port Python support to Python 3

• Entities are no longer regarded as defined due to being part of a cleared transaction. --explicit is effectively enabled by default and is now a no-op (PR #1819)

• Add --average-lot-prices to show the average of lot prices

• Add support for %F date format specifier (#1775)

• Add commodity_price(NAME, DATE) function

• Add set_commodity_price(NAME, DATE) function

• Fix buffer overflow when evaluating date

• Fix balance assertions on accounts with virtual posts (#543)

• Fix segfault with ledger print (#1850)

• Ensure that apply directives (like apply account) have the required argument (#553)

• Format annotations using a date format that can be parsed

• Change --invert to invert displayed amounts and totals, not amounts (#1803)

• Correct the way parens are parsed in query expressions

• Compare price annotations using their textual rendering

• Fix build failure with utfcpp 3.0 (#1816)

• Fix build failure due to ambiguous type (#1833)

• Properly reject postings with a comment right after the flag (#1753)

• Make sorting order of lot information deterministic (#1747)

• Fix bug in tag value parsing (#1702)

• Remove the org command, which was always a hack to begin with (#1706)

• Provide Docker information in README

• Various small documentation improvements

• Increase maximum length for regex from 255 to 4095 (#981)

• Initialize periods from from/since clause rather than earliest transaction date (#1159)

• Check balance assertions against the amount after the posting (#1147)

• Allow balance assertions with multiple posts to same account (#1187)

• Fix period duration of "every X days" and similar statements (#370)

• Make option --force-color not require --color anymore (#1109)

• Add quoted_rfc4180 to allow CVS output with RFC 4180 compliant quoting.

• Add support for --prepend-format in accounts command

• Fix handling of edge cases in trim function (#520)

• Fix auto xact posts not getting applied to account total during journal parse (#552)

• Transfer null_post flags to generated postings

• Fix segfault when using --market with --group-by

• Use amount_width variable for budget report

• Keep pending items in budgets until the last day they apply

• Fix bug where .total used in value expressions breaks totals

• Make automated transactions work with assertions (#1127)

• Improve parsing of date tokens (#1626)

• Don't attempt to invert a value if it's already zero (#1703)

• Do not parse user-specified init-file twice

• Fix parsing issue of effective dates (#1722, TALOS-2017-0303, CVE-2017-2807)

• Fix use-after-free issue with deferred postings (#1723, TALOS-2017-0304, CVE-2017-2808)

• Fix possible stack overflow in option parsing routine (#1222, CVE-2017-12481)

• Fix possible stack overflow in date parsing routine (#1224, CVE-2017-12482)

• Fix use-after-free when using --gain (#541)

• Python: Removed double quotes from Unicode values.

• Python: Ensure that parse errors produce useful RuntimeErrors

• Python: Expose journal expand_aliases

• Python: Expose journal_t::register_account

• Improve bash completion

• Emacs Lisp files have been moved to https://github.com/ledger/ledger-mode

• Fix build under MSYS (32-bit).

• Fix build under Cygwin.

• Various documentation improvements

• Added a --no-revalued option

• Improved Embedded Python Support

• Use ./.ledgerrc if ~/.ledgerrc doesn't exist

• Fixed parsing of transactions with single-character payees and comments

• Fixed crash when using -M with empty result

• Fixed sorting for option --auto-match

• Fixed treatment of year 2015 and Y2014 directives

• Fixed crash when using --trace 10 or above

• Build fix for boost 1.58, 1.59, 1.60

• Build fix for Cygwin

• Fixed Util and Math tests on Mac OS X

• Various documentation improvements

• Examples in the documentation are tested just like unit tests

• Add continuous integration (https://travis-ci.org/ledger/ledger)

• Changed the definition of cost basis to preserve the original cost basis when a gain or loss is made (if you bought 1 AAA for $10 and then sold it for $12, ledger would previously take $12 as the cost; the original cost of $10 is preserved as the cost basis now, which addresses strange behavior with -B after a capital gain or loss is made).

• Incorrect automatic Equity:Capital Gains and Equity:Capital Loss entries are no longer generated when a commodity is sold for loss or profit.

• Support for virtual posting costs.

• The option --permissive now quiets balance assertions

• Removed SHA1 files due to license issues and use boost instead.

• Added option --no-pager to disable the pager.

• Added option --no-aliases to completely disable alias expansion

• Added option --recursive-aliases to expand aliases recursively

• Support payee uuid directive.

• Bug fix: when a status flag (! or *) is explicitly specified for an individual posting, it always has a priority over entire transaction status.

• Bug fix: don't lose commodity when cost is not separated by whitespace

• Improved backwards compatibility with ledger 2.x

• Build fix for GCC 4.9

• Build fix for boost 1.56

• Many improvements to ledger-mode, including fontification

• More test cases and unit tests

• Contrib: Added script to generate commodities from ISO 4217

Due to the magnitude of changes in 3.0, only changes that affect compatibility with 2.x files and usage is mentioned here. For a description of new features, please see the manual.

• The option -g (--performance) was removed.

• The balance report now defaults to showing all relevant accounts. This is the opposite of 2.x. That is, bal in 3.0 does what -s bal did in 2.x. To see 2.6 behavior, use bal -n in 3.0. The -s option no longer has any effect on balance reports.

• Minor fixes to allow for compilation with gcc 4.4.

• Bug fix: Command-line options, such as -O, now override init-file options such as -V.

• Bug fix: "cat data | ledger -f -" now works.

• Bug fix: --no-cache is now honored. Previously, it was writing out a cache file named "".

• Bug fix: Using %.2X in a format string now outputs 2 spaces if the state is cleared.

• Added the concept of "balance setting transactions":

  Setting an account's balance

  You can now manually set an account's balance to whatever you want, at any time. Here's how it might look at the beginning of your Ledger file:

  2008/07/27 Starting fresh
      Assets:Checking      = $1,000.00
      Equity:Opening Balances
  

  If Assets:Checking is empty, this is no different from omitting the "=". However, if Assets:Checking did have a prior balance, the amount of the transaction will be auto-calculated so that the final balance of Assets:Checking is now $1,000.00.

  Let me give an example of this. Say you have this:

  2008/07/27 Starting fresh
      Assets:Checking          $750.00
      Equity:Opening Balances
  
  2008/07/27 Starting fresh
      Assets:Checking      = $1,000.00
      Equity:Adjustments
  

  These two entries are exactly equivalent to these two:

  2008/07/27 Starting fresh
      Assets:Checking          $750.00
      Equity:Opening Balances
  
  2008/07/27 Starting fresh
      Assets:Checking          $250.00
      Equity:Adjustments
  

  The use of the "=" sign here is that it sets the transaction's amount to whatever is required to satisfy the assignment. This is the behavior if the transaction's amount is left empty.

  Multiple commodities

  As far as commodities go, the = sign only works if the account balance's commodity matches the commodity of the amount after the equals sign. However, if the account has multiple commodities, only the matching commodity is affected. Here's what I mean:

  2008/07/24 Opening Balance
      Assets:Checking        = $250.00          ; we force set it
      Equity:Opening Balances
  
  2008/07/24 Opening Balance
      Assets:Checking      = EC 250.00          ; we force set it again
      Equity:Opening Balances
  

  This is an error, because $250.00 cannot be auto-balanced to match EC 250.00. However:

  2008/07/24 Opening Balance
      Assets:Checking        = $250.00          ; we force set it again
      Assets:Checking        EC 100.00          ; and add some EC's
      Equity:Opening Balances
  
  2008/07/24 Opening Balance
      Assets:Checking      = EC 250.00          ; we force set the EC's
      Equity:Opening Balances
  

  This is not an error, because the latter auto-balancing transaction only affects the EC 100.00 part of the account's balance; the $250.00 part is left alone.

  Checking statement balances

  When you reconcile a statement, there are typically one or more transactions which result in a known balance. Here's how you specify that in your Ledger data:

  2008/07/24 Opening Balance
      Assets:Checking        = $100.00
      Equity:Opening Balances
  
  2008/07/30 We spend money, with a known balance afterward
      Expenses:Food             $20.00
      Assets:Checking         = $80.00
  
  2008/07/30 Again we spend money, but this time with all the info
      Expenses:Food             $20.00
      Assets:Checking          $-20.00 = $60.00
  
  2008/07/30 This entry yield an 'unbalanced' error
      Expenses:Food             $20.00
      Assets:Checking          $-20.00 = $30.00
  

  The last entry in this set fails to balance with an unbalanced remainder of $-10.00. Either the entry must be corrected, or you can have Ledger deal with the remainder automatically:

  2008/07/30 The fixed entry
      Expenses:Food              $20.00
      Assets:Checking           $-20.00 = $30.00
      Equity:Adjustments
  

  Conclusion

  This simple feature has all the utility of @check, plus auto-balancing to match known target balances, plus the ability to guarantee that an account which uses only one commodity does contain only that commodity.

  This feature slows down textual parsing slightly, does not affect speed when loading from the binary cache.

• The rest of the changes in the version is all bug fixes (around 45 of them).

• Gnucash parser is fixed.

• Fix a memory leak bug in the amount parser.

• (This feature is from 2.6, but was not documented anywhere):

  Commodities may now specify lot details, to assign in managing set groups of items, like buying and selling shares of stock.

  For example, let's say you buy 50 shares of AAPL at $10 a share:

  2007/01/14 Stock purchase
      Assets:Brokerage         50 AAPL @ $10
      Assets:Brokerage
  

  Three months later, you sell this "lot". Based on the original purchase information, Ledger remembers how much each share was purchased for, and the date on which it was purchased. This means you can sell this specific lot by price, by date, or by both. Let's sell it by price, this time for $20 a share.

  2007/04/14 Stock purchase
      Assets:Brokerage         $1000.00
      Assets:Brokerage         -50 AAPL {$10} @ $20
      Income:Capital Gains     $-500.00
  

  Note that the Income:Capital Gains line is now required to balance the transaction. Because you sold 50 AAPL at $20/share, and because you are selling shares that were originally valued at $10/share, Ledger needs to know how you will "balance" this difference. An equivalent Expenses:Capital Loss would be needed if the selling price were less than the buying price.

  Here's the same example, this time selling by date and price:

  2007/04/14 Stock purchase
      Assets:Brokerage         $1000.00
      Assets:Brokerage         -50 AAPL {$10} [2007/01/14] @ $20
      Income:Capital Gains     $-500.00
  

  If you attempt to sell shares for a date you did not buy them, note that Ledger will not complain (as it never complains about the movement of commodities between accounts). In this case, it will simply create a negative balance for such shares within your Brokerage account; it's up to you to determine whether you have them or not.

• To facilitate lot pricing reports, there are some new reporting options:

  • --lot-prices Report commodities with different lot prices as if they were different commodities. Otherwise, Ledger just gloms all the AAPL shares together.

  • --lot-dates Separate commodities by lot date. Every transaction that uses the '@' cost specifier will have an implicit lot date and lot price.

  • --lot-tags Separate commodities by their arbitrary note tag. Note tags may be specified using (note) after the commodity.

  • --lots Separate commodities using all lot information.

• The style for eliding long account names (for example, in the register report) has been changed. Previously Ledger would elide the end of long names, replacing the excess length with "..". However, in some cases this caused the base account name to be missing from the report!

  What Ledger now does is that if an account name is too long, it will start abbreviating the first parts of the account name down to two letters in length. If this results in a string that is still too long, the front will be elided -- not the end. For example:

  Expenses:Cash           ; OK, not too long
  Ex:Wednesday:Cash       ; "Expenses" was abbreviated to fit
  Ex:We:Afternoon:Cash    ; "Expenses" and "Wednesday" abbreviated
  ; Expenses:Wednesday:Afternoon:Lunch:Snack:Candy:Chocolate:Cash
  ..:Af:Lu:Sn:Ca:Ch:Cash  ; Abbreviated and elided!
  

  As you can see, it now takes a very deep account name before any elision will occur, whereas in 2.x elisions were fairly common.

• In addition to the new elision change mentioned above, the style is also configurable:

  • --truncate leading ; elide at the beginning

  • --truncate middle ; elide in the middle

  • --truncate trailing ; elide at end (Ledger 2.x's behavior)

  • --truncate abbrev ; the new behavior

  • --abbrev-len 2 ; set length of abbreviations

  These elision styles affect all format strings which have a maximum width, so they will also affect the payee in a register report, for example. In the case of non-account names, "abbrev" is equivalent to "trailing", even though it elides at the beginning for long account names.

• Error reporting has been greatly improving, now showing full contextual information for most error messages.

• Added --base reporting option, for reporting convertible commodities in their most basic form. For example, if you read a timeclock file with Ledger, the time values are reported as hour and minutes -- whichever is the most compact form. But with --base, Ledger reports only in seconds.

  NOTE: Setting up convertible commodities is easy; here's how to use Ledger for tracking quantities of data, where the most compact form is reported (unless --base is specified):

  C 1.00 Kb = 1024 b
  C 1.00 Mb = 1024 Kb
  C 1.00 Gb = 1024 Mb
  C 1.00 Tb = 1024 Gb
  

• Added --ansi reporting option, which shows negative values in the running total column of the register report as red, using ANSI terminal codes; --ansi-invert makes non-negative values red (which makes more sense for the income and budget reports).

  The --ansi functionality is triggered by the format modifier "!", for example the register reports uses the following for the total (last) column:

  %!12.80T
  

  At the moment neither the balance report nor any of the other reports make use of the ! modifier, and so will not change color even if --ansi is used. However, you can modify these report format strings yourself in ~/.ledgerrc if you wish to see red coloring of negative sums in other places.

• Added --only predicate, which occurs during transaction processing between --limit and --display. Here is a summary of how the three supported predicates are used:

  • --limit "a>100"

    This flag limits computation to only transactions whose amount is greater than 100 of a given commodity. It means that if you scan your dining expenses, for example, only individual bills greater than $100 would be calculated by the report.

  • --only "a>100"

    This flag happens much later than --limit, and corresponding more directly to what one normally expects. If --limit isn't used, then ALL your dining expenses contribute to the report, but only those calculated transactions whose value is greater than $100 are used. This becomes important when doing a monthly costs report, for example, because it makes the following command possible:

    ledger -M --only "a>100" reg ^Expenses:Food
    

    This shows only months whose amount is greater than 100. If --limit had been used, it would have been a monthly summary of all individual dinner bills greater than 100 -- which is a very different thing.

  • --display "a>100"

    This predicate does not constrain calculation, but only display. Consider the same command as above:

      ledger -M --display "a>100" reg ^Expenses:Food
    

    This displays only lines whose amount is greater than 100, yet the running total still includes amounts from all transactions. This command has more particular application, such as showing the current month's checking register while still giving a correct ending balance:

      ledger --display "d>[this month]" reg Checking
    

    Note that these predicates can be combined. Here is a report that considers only food bills whose individual cost is greater than $20, but shows the monthly total only if it is greater than $500. Finally, we only display the months of the last year, but we retain an accurate running total with respect to the entire ledger file:

      ledger -M --limit "a>20" --only "a>200" \
        --display "year == yearof([last year])" reg ^Expenses:Food
    

• Added new "--descend AMOUNT" and "--descend-if VALEXPR" reporting options. For any reports that display valued transactions (i.e., register, print, etc), you can now descend into the component transactions that made up any of the values you see.

  For example, say you request a --monthly expenses report:

  $ ledger --monthly register ^Expenses
  

  Now, in one of the reported months you see $500.00 spent on Expenses:Food. You can ask Ledger to "descend" into, and show the component transactions of, that $500.00 by respecifying the query with the --descend option:

  $ ledger --monthly --descend "\$500.00" register ^Expenses
  

  The --descend-if option has the same effect, but takes a value expression which is evaluated as a boolean to locate the desired reported transaction.

• Added a "dump" command for creating binary files, which load much faster than their textual originals. For example:

  ledger -f huge.dat -o huge.cache dump
  ledger -f huge.cache bal
  

  The second command will load significantly faster (usually about six times on my machine).

• There have a few changes to value expression syntax. The most significant incompatibilities being:

  • Equality is now ==, not =
  • The U, A, and S functions now requires parens around the argument. Whereas before At was acceptable, now it must be specified as A(t).
  • The P function now always requires two arguments. The old one-argument version P(x) is now the same as P(x,m).

  The following value expression features are new:

  • A C-like comma operator is supported, where all but the last term are ignored. The is significant for the next feature:

  • Function definitions are now supported. Scoping is governed by parentheses. For example: (x=100, x+10) ; yields 110 as the result (f(x)=x*2,f(100)) ; yields 200 as the result

  • Identifier names may be any length. Along with this support comes alternate, longer names for all of the current one-letter value expression variables:

    Old New

    ---

    m now a amount a amount b cost i price d date X cleared Y pending R real L actual n index N count l depth O total B cost_total I price_total v market V market_total g gain G gain_total U(x) abs(x) S(x) quant(x), quantity(x) comm(x), commodity(x) setcomm(x,y), set_commodity(x,y) A(x) mean(x), avg(x), average(x) P(x,y) val(x,y), value(x,y) min(x,y) max(x,y)

• There are new "parse" and "expr" commands, whose argument is a single value expression. Ledger will simply print out the result of evaluating it. "parse" happens before parsing your ledger file, while "expr" happens afterward. Although "expr" is slower as a result, any commodities you use will be formatted based on patterns of usage seen in your ledger file.

  These commands can be used to test value expressions, or for doing calculation of commoditized amounts from a script.

  A new "--debug" will also dump the resulting parse tree, useful for submitting bug reports.

• Added new min(x,y) and max(x,y) value expression functions.

• Value expression function may now be defined within your ledger file (or initialization file) using the following syntax:

  @def foo(x)=x*1000
  

  This line makes the function "foo" available to all subsequent value expressions, to all command-line options taking a value expression, and to the new "expr" command (see above).

• Added a new value expression regexp command: C// compare against a transaction amount's commodity symbol

• Added a new "csv" command, for outputting results in CSV format.

• Ledger now expands ~ in file pathnames specified in environment variables, initialization files and journal files.

• Effective dates may now be specified for entries:

  2004/10/03=2004/09/30 Credit card company
      Liabilities:MasterCard         $100.00
      Assets:Checking
  

  This entry says that although the actual transactions occurred on October 3rd, their effective date was September 30th. This is especially useful for budgeting, in case you want the transactions to show up in September instead of October.

  To report using effective dates, use the --effective option.

• Actual and effective dates may now be specified for individual transactions:

  2004/10/03=2004/09/30 Credit card company
      Liabilities:MasterCard         $100.00
      Assets:Checking                         ; [2004/10/10=2004/09/15]
  

  This states that although the actual date of the entry is 2004/10/03, and the effective date of the entry is 2004/09/30, the actual date of the Checking transaction itself is 2004/10/10, and its effective date is 2004/09/15. The effective date is optional (just specifying the actual date would have read "[2004/10/10]").

  If no effective date is given for a transaction, the effective date of the entry is assumed. If no actual date is given, the actual date of the entry is assumed. The syntax of the latter is simply [=2004/09/15].

• To support the above, there is a new formatting option: "%d". This outputs only the date (like "%D") if there is no effective date, but outputs "ADATE=EDATE" if there is one. The "print" report now uses this.

• To support the above, the register report may now split up entries whose component transactions have different dates. For example, given the following entry:

  2005/10/15=2005/09/01 iTunes
      Expenses:Music                 $1.08 ; [2005/10/20=2005/08/01]
      Liabilities:MasterCard
  

  The command "ledger register" on this data file reports:

  2005/10/20 iTunes   Expenses:Music            $1.08    $1.08
  2005/10/15 iTunes   Liabilities:MasterCard   $-1.08        0
  

  While the command "ledger --effective register" reports:

  2005/08/01 iTunes   Expenses:Music            $1.08    $1.08
  2005/09/01 iTunes   Liabilities:MasterCard   $-1.08        0
  

  Although it appears as though two entries are being reported, both transactions belong to the same entry.

• Individual transactions may now be cleared separately. The old syntax, which is still supported, clears all transactions in an entry:

  2004/05/27 * Book Store
      Expenses:Dining                 $20.00
      Liabilities:MasterCard
  

  The new syntax allows clearing of just the MasterCard transaction:

  2004/05/27 Book Store
      Expenses:Dining                 $20.00
      * Liabilities:MasterCard
  

  NOTE: This changes the output format of both the "emacs" and "xml" reports. ledger.el uses the new syntax unless the Lisp variable `ledger-clear-whole-entries' is set to t.

• Removed Python integration support.

• Did much internal restructuring to allow the use of libledger.so in non-command-line environments (such as GUI tools).

• Corrected an issue that had inadvertently disabled Gnucash support.

• Both -$100.00 and $-100.00 are now equivalent amounts.

• Simple, inline math (using the operators +-/*, and/or parentheses) is supported in transactions. For example:

  2004/05/27 Book Store
      Expenses:Dining                 $20.00 + $2.50
      Liabilities:MasterCard
  

  This won't register the tax/tip in its own account, but might make later reading of the ledger file easier.

• Use of a "catch all" account is now possible, which auto-balances entries that contain only one transaction. For sanity's sake this is not used to balance all entries, as that would make locating unbalanced entries a nightmare. Example:

  A Liabilities:MasterCard
  
  2004/05/27 Book Store
      Expenses:Dining                 $20.00 + $2.50
  

  This is equivalent to the entry in the previous bullet.

• Entries that contain a single transaction with no amount now always balance, even if multiple commodities are involved. This means that the following is now supported, which wasn't previously:

  2004/06/21 Adjustment
      Retirement          100 FUNDA
      Retirement          200 FUNDB
      Retirement          300 FUNDC
      Equity:Adjustments
  

• Fixed several bugs relating to QIF parsing, budgeting and forecasting.

• The configure process now looks for libexpat in addition to searching for libxmlparse+libxmltok (how expat used to be packaged).

• The directive "!alias ALIAS = ACCOUNT" makes it possible to use "ALIAS" as an alternative name for ACCOUNT in a textual ledger file. You might use this to associate the single word "Bank" with the checking account you use most, for example.

• The --version page shows the optional modules ledger was built with.

• Fixed several minor problems, plus a few major ones dealing with imprecise date parsing.

• Ledger now compiles under gcc 2.95.

• Fixed several core engine bugs, and problems with Ledger's XML data format.

• Errors in XML or Gnucash data now report the correct line number for the error, instead of always showing line 1.

• 'configure' has been changed to always use a combination of both compile and link tests for every feature, in order to identify environment problems right away.

• The "D " command, released in 2.1, now requires a commoditized amount, such as "D $1,000.00". This sets not only the default commodity, but several flags to be used with all such commodities (such as whether numbering should be American or European by default). This entry may be used be many times; the most recent seen specifies the default for entries that follow.

• The binary cache now remembers the price history database that was used, so that if LEDGER_PRICE_DB is silently changed, the cache will be thrown away and rebuilt.

• OFX data importing is now supported, using libofx (http://libofx.sourceforge.net). configure will check if the library is available. You may need to add CPPFLAGS or LDFLAGS to the command-line for the appropriate headers and library to be found. This support is preliminary, and as such is not documented yet.

• All journal entries now remember where they were read from. New format codes to access this information are: %S for source path, %B for beginning character position, and %E for ending character position.

• Added "pricesdb" command, which is identical to "prices" except that it uses the same format as Ledger's usual price history database.

• Added "output FILE" command, which attempts to reproduce the input journal FILE exactly. Meant for future GUI usage. This command relies on --write-hdr-format and --write-xact-format, instead of --print-format.

• Added "--reconcile BALANCE" option, which attempts to reconcile all matching transactions to the given BALANCE, outputting those that would need to be "cleared" to match it. Using by the auto-reconciling feature of ledger.el (see below).

  "--reconcile-date DATE" ignores any uncleared transactions after DATE in the reconciling algorithm. Since the algorithm is O(n^2) (where 'n' is the number of uncleared transactions to consider), this could have a substantial impact.

• In ledger.el's Reconcile mode ('C-c C-r' from a ledger-mode file):

  • 'a' adds a missing transaction
  • 'd' deletes the current transaction
  • 'r' attempts to auto-reconcile (same as 'C-u C-c C-r')
  • 's' or 'C-x C-s' will save the ledger data file and show the currently cleared balance
  • 'C-c C-c' commits the pending transactions, marking them cleared.

  This feature now works with Emacs 21.3. Also, the reconciler no longer needs to ask "how far back" to go.

• To support the reconciler, textual entries may now have a "!" flag (pending) after the date, instead of a "*" flag (cleared).

• There are a new set of value expression regexp commands:

  • c// entry code
  • p// payee
  • w// short account name
  • W// full account name
  • e// transaction note

  This makes it possible to display transactions whose comment field matches a particular text string. For example:

  ledger -l e/{tax}/ reg
  

  prints out all the transactions with the comment "{tax}", which might be used to identify items related to a tax report.

• Improved the autoconf system to be smarter about finding XML libs

• Added --no-cache option, to always ignore any binary cache file

• `ledger-reconcile' (in ledger.el) no longer asks for a number of days

• Fixed %.XY format, where X is shorter than the string generated by Y

• New directive for text files: "D " specifies the default commodity used by the entry command

This version represents a full rewrite, while preserving much of the original data format and command-line syntax. There are too many new features to describe in full, but a quick list: value expressions, complex date masks, binary caching of ledger data, several new reporting options, a simple way to specify payee regexps, calculation and display predicates, and two-way Python integration. Ledger also uses autoconf now, and builds as a library in addition to a command-line driver.

• changes in option syntax:

  -d now specifies the display predicate. To give a date mask similar to 1.7, use the -p (period) option.

  -P now generates the "by payee" report. To specify a price database to use, use --price-db.

  -G now generates a net gain report. To print totals in a format consumable by gnuplot, use -J.

  -l now specifies the calculation predicate. To emulate the old usage of "-l $100", use: -d "AT>100".

  -N is gone. Instead of "-N REGEX", use: -d "/REGEX/?T>0:T".

  -F now specifies the report format string. The old meaning of -F now has little use.

  -S now takes a value expression as the sorting criterion. To get the old meaning of "-S", use "-S d".

  -n now means "collapse entries in the register report". The get the old meaning of -n in the balance report, use "-T a".

  -p now specifies the reporting period. You can convert commodities in a report using value expressions. For example, to display hours at $10 per hour:

  -T "O>={0.01h}?{\$10.00}*O:O"
  

  Or, to reduce totals, so that every $417 becomes 1.0 AU:

  -T "O>={\$0.01}?{1.0 AU}*(O/{\$417}):O"
  

• The use of "+" and "-" in ledger files to specify permanent regexps has been removed.

• The "-from" argument is no longer used by the "entry" command. Simply remove it.

• The most significant feature to be added is "value expressions". They are used in many places to indicate what to display, sorting order, how to calculate totals, etc. Logic and math operators are supported, as well as simple functions. See the manual.

• If the environment variable LEDGER_FILE (or LEDGER) is used, a binary cache of that ledger is kept in ~/.ledger-cache (or the file given by LEDGER_CACHE). This greatly speeds up subsequent queries. Happens only if "-f" or "--file" is not used.

• New 'xml' report outputs an XML version of what "register" would have displayed. This can be used to manipulate reported data in a more scriptable way.

  Ledger can also read as input the output from the "xml" report. If the "xml" report did not contain balanced entries, they will be balanced by the "" account. For example:

  ledger reg rent
  

  displays the same results as:

  ledger xml rent | ledger -f - reg rent
  

• Regexps given directly after the command name now apply only to account names. To match on a payee, use "--" to separate the two kinds of regexps. For example, to find a payee named "John" within all Expenses accounts, use:

  ledger register expenses -- john
  

  Note: This command is identical (and internally converted) to:

  ledger -l "/expenses/|//john/" register
  

• To include entries from another file into a specific account, use:

  !account ACCOUNT
  !include FILE
  !end
  

• Register reports now show only matching account transactions. Use "-r" to see "related accounts" -- the account the transfer came from or went to (This was the old behavior in 1.x, but led to confusion). "-r" also works with balance reports, where it will total all the transactions related to your query.

• Automated transactions now use value expressions for the predicate. The new syntax is:

  = VALUE-EXPR
    TRANSACTIONS...
  

  Only one VALUE-EXPR is supported (compared to multiple account regexps before). However, since value expression allow for logic chaining, there is no loss of functionality. Matching can also be much more comprehensive.

• If Boost.Python is installed (libboost_python.a), ledger can support two-way Python integration. This feature is enabled by passing --enable-python to the "configure" script before building. Ledger can then be used as a module (ledger.so), as well as supporting Python function calls directly from value expressions. See main.py for an example of driving Ledger from Python. It implements nearly all the functionality of the C++ driver, main.cc.

  (This feature has yet to mature, and so is being offered as a beta feature in this release. It is mostly functional, and those curious are welcome to play with it.)

• New reporting options:

  • "-o FILE" outputs data to FILE. If "-", output goes to stdout (the default).

  • -O shows base commodity values (this is the old behavior)

  • -B shows basis cost of commodities

  • -V shows market value of commodities

  • -g reports gain/loss performance of each register item

  • -G reports net gain/loss over time

  • -A reports average transaction value (arithmetic mean)

  • -D reports each transaction's deviation from the average

  • -w uses 132 columns for the register report, rather than 80. Set the environment variable LEDGER_WIDE for this to be the default.

  • "-p INTERVAL" allows for more flexible period reporting, such as:

    monthly every week every 3 quarters weekly from 12/20 monthly in 2003 weekly from last month until dec

  • "-y DATEFMT" changes the date format used in all reports. The default is "%Y/%m/%d".

    -Y and -W print yearly and weekly subtotals, just as -M prints monthly subtotals.

  • --dow shows cumulative totals for each day of the week.

  • -P reports transactions grouped by payee

  • -x reports the payee as the commodity; useful in some cases

  • -j and -J replace the previous -G (gnuplot) option. -j reports the amounts column in a way gnuplot can consume, and -J the totals column. An example is in "scripts/report".

  • "--period-sort EXPR" sorts transactions within a reporting period. The regular -S option sorts all reported transactions.

• Pricing histories are now supported, so that ledger remembers the historical prices of all commodities, and can present register reports based on past and present market values as well as original cost basis. See the manual for more details on the new option switches.

• Ledger can now parse timeclock files. These are simple timelogs that track in/out events, which can be maintained using my timeclock tool. By allowing ledger to parse these, it means that reporting can be done on them in the same way as ledger files (the commodity used is "h", for hours); it means that doing things like tracking billable hours for clients, and invoicing those clients to transfer hours into dollar values via a receivable account, is now trivial. See the docs for more on how to do this.

• Began keeping a NEWS file. :)