;doc: move more content out of CONTRIBUTING

2021-09-27 01:17:18 -10:00 · 2021-09-27 01:17:18 -10:00 · 0fc4181456
commit 0fc4181456
parent fd052d308b
6 changed files with 583 additions and 543 deletions
--- a/CHANGELOGS.md
+++ b/CHANGELOGS.md
@ -1,8 +1,28 @@
 # Changelogs
-Always maintain changelogs in master branch (not in release branches).
+<div class=pagetoc>
 <!-- toc -->
 </div>
-**Frequently**, especially after merging changes, and before cherry picking into release branch:
+## How to prepare changelogs & release notes
 Changelogs:
 - always maintain changelogs in master branch, not in release branches
 - ./Shake changelogs
 - edit the new changelog items (identify, filter, move to correct changelog, deduplicate, rewrite, sort/group)
 Release notes:
 - add a new TOC entry and section in site/release-notes.md
 - copy/rewrite/summarise package changelogs 
 - note any other items of interest
 - list release contributors
 - write release summary
 ## Frequently
 especially after merging changes, and before cherry picking into release branch:
 - dry run: `./Shake changelogs -n`
 - add new changes: `./Shake changelogs`
@ -14,7 +34,10 @@ Always maintain changelogs in master branch (not in release branches).
  - format ([#ISSUE](https://github.com/simonmichael/hledger/issues/), AUTHOR) on its own line
 - commit: `./Shake changelogs -c`
-**After cherry-picking** changes to a release branch:
+## After cherry-picking
 changes to a release branch:
 - in the master branch changelogs, move the corresponding changelog items under a pending release heading,
  creating that when necessary:
    ```
@ -29,7 +52,7 @@ Always maintain changelogs in master branch (not in release branches).
    # LASTVER YYYY-MM-DD
    ```
-**At release:**
+## At release
 - do final update/edits; check organisation, wording, formatting, issue links
 - replace "unreleased" with the release date
--- a/COMMITS.md
+++ b/COMMITS.md
@ -1,8 +1,12 @@
 # COMMITS
 <div class=pagetoc>
 <!-- toc -->
 </div>
 ## When committing / reviewing commits
-Follow/encourage the [commit conventions](CONTRIBUTING.html#commit-messages).
+Follow/encourage the commit conventions (see below).
 Here they are in brief:
 -  Commit messages must begin with a prefix, one or more colon-terminated words
   indicating the [topic](CONTRIBUTING.html#topics).
@ -16,7 +20,96 @@ Here they are in brief:
   what they need to know.
 ## When committing / pushing / merging:
 - run `bin/commitlint` before push, to check recent commits
 - or, run it automatically before each commit (`make installcommithook` to configure your local repo)
 - it also runs in CI on github for pull requests, etc.
 ## Commit conventions
 Since the 1.23 release cycle, we try to follow certain conventions for commit messages, to
 - encourage considered, focussed, well documented changes
 - reduce the cost of code review, maintaining changelogs and release notes, and releasing
 - increase our throughput (rate of shipping useful, reliable, documented, maintainable features)
 **hledger commit conventions:**
 1. Commit messages in hledger's main repo follow this pattern:
   ```
   type: [optionaltopic:] summary
   [Optional description, more details here when needed.]
   ```
 2. Every top-level commit must have a type prefix, ending with a colon and optional space. 
   This indicates the change's intended audience and the general type of change. 
   Here are the current types:
   - **Changes visible to end users** (including users of hledger-web's HTTP API).
     These will appear in release notes and changelogs:
     - `feat` - a new feature
     - `imp`  - an improvement to existing features
     - `fix`  - a bugfix
   - **Changes affecting packagers, builders, and library users**. 
    These will appear in changelogs:
     - `cha` - a generic package/lib change. Or, one of these specific types:
     - `pkg` - something to do with the haskell packages, dependencies etc.
     - `lib` - a change in the package's library API
     - ...some other type that seems useful...
   - **Changes interesting only to hledger developers/documentors/debuggers**.
     These will usually appear only in the commit history, not in changelogs or release notes:
     - `dev` - a generic developer change. Or, one of these specific types:
     - `ref` - refactoring
     - `cln` - cleanup
     - `doc` - documentation-related
     - `test` - tests-related
     - `ci`  - continuous integration-related
     - ...some other type that seems useful...
    There's a bit of ambiguity/overlap between the cha/dev types and topics.
    Eg the `doc` type indicates a boring doc change, but there's also a `doc` topic
    which might be used for interesting doc changes, as in `feat:doc:...`. TBD.
 3. If this is a "breaking change", introducing a compatibility or
   migration issue, the type is followed by `!`, and the issue
   and advice to users are included in the description.
   This will most often be seen with the end-user types, eg:
   `feat!:`, `imp!:`, `fix!:`.
 4. If the first character of the commit message is `;`, this commit
   (more precisely, the push ending with this commit) will be excluded
   from the usual CI checks. Our CI tends to do a lot of building, so 
   you can use this to save energy and carbon emissions when pushing 
   harmless changes.
 5. A topic prefix, and maybe even a subtopic prefix, can be added
   before the summary if useful. These are standard prefixes similar
   to what I have been using for some time, see [topics](#topics). 
   They help with readability in the commit history, changelogs and release notes.
 6. Any relevant issues should be mentioned, often parenthesised at 
   the end of the summary: `(#NNNN)`.
 7. The summary, and description if any, communicate this change's
   purpose as clearly as possible to its intended audience: 
   end users, builders/packagers/library users, or developers/debuggers.
   The text (or its first sentence/first paragraphs) should be ready 
   for use in changelogs/release notes when applicable.
 Crafting good commit messages (and thereby good commits, good change
 documentation, easier code review, faster merging) is an art and a
 habit. Just use your best judgement and we'll check and polish
 as part of CI and code review. Examples will be added here in due course.
 Related:
 - <https://groups.google.com/g/hledger/c/t14xwHQMJqU/m/9frARXIdAAAJ>
 - <https://conventionalcommits.org>
 - <https://git.savannah.gnu.org/cgit/emacs.git/plain/CONTRIBUTE> -> Commit messages
--- a/CONTRIBUTING.md
+++ b/CONTRIBUTING.md
@ -4,6 +4,10 @@
 <!-- toc -->
 </div>
 This is a collection of old developer docs on all topics.
 Gradually, these are moving to separate files/pages
 and this doc is becoming a focussed guide for new contributors.
 ## Quick Links
 | | |
@ -503,211 +507,6 @@ Relevant tools include:
  [log](https://ircbrowse.net/day/hledger/2015/10/11) -->
 ## Commit messages
 Since the 1.23 release cycle, we try to follow certain conventions for commit messages, to
 - encourage considered, focussed, well documented changes
 - reduce the cost of code review, maintaining changelogs and release notes, and releasing
 - increase our throughput (rate of shipping useful, reliable, documented, maintainable features)
 **hledger commit conventions:**
 1. Commit messages in hledger's main repo follow this pattern:
   ```
   type: [optionaltopic:] summary
   [Optional description, more details here when needed.]
   ```
 2. Every top-level commit must have a type prefix, ending with a colon and optional space. 
   This indicates the change's intended audience and the general type of change. 
   Here are the current types:
   - **Changes visible to end users** (including users of hledger-web's HTTP API).
     These will appear in release notes and changelogs:
     - `feat` - a new feature
     - `imp`  - an improvement to existing features
     - `fix`  - a bugfix
   - **Changes affecting packagers, builders, and library users**. 
    These will appear in changelogs:
     - `cha` - a generic package/lib change. Or, one of these specific types:
     - `pkg` - something to do with the haskell packages, dependencies etc.
     - `lib` - a change in the package's library API
     - ...some other type that seems useful...
   - **Changes interesting only to hledger developers/documentors/debuggers**.
     These will usually appear only in the commit history, not in changelogs or release notes:
     - `dev` - a generic developer change. Or, one of these specific types:
     - `ref` - refactoring
     - `cln` - cleanup
     - `doc` - documentation-related
     - `test` - tests-related
     - `ci`  - continuous integration-related
     - ...some other type that seems useful...
    There's a bit of ambiguity/overlap between the cha/dev types and topics.
    Eg the `doc` type indicates a boring doc change, but there's also a `doc` topic
    which might be used for interesting doc changes, as in `feat:doc:...`. TBD.
 3. If this is a "breaking change", introducing a compatibility or
   migration issue, the type is followed by `!`, and the issue
   and advice to users are included in the description.
   This will most often be seen with the end-user types, eg:
   `feat!:`, `imp!:`, `fix!:`.
 4. If the first character of the commit message is `;`, this commit
   (more precisely, the push ending with this commit) will be excluded
   from the usual CI checks. Our CI tends to do a lot of building, so 
   you can use this to save energy and carbon emissions when pushing 
   harmless changes.
 5. A topic prefix, and maybe even a subtopic prefix, can be added
   before the summary if useful. These are standard prefixes similar
   to what I have been using for some time, see [topics](#topics). 
   They help with readability in the commit history, changelogs and release notes.
 6. Any relevant issues should be mentioned, often parenthesised at 
   the end of the summary: `(#NNNN)`.
 7. The summary, and description if any, communicate this change's
   purpose as clearly as possible to its intended audience: 
   end users, builders/packagers/library users, or developers/debuggers.
   The text (or its first sentence/first paragraphs) should be ready 
   for use in changelogs/release notes when applicable.
 Crafting good commit messages (and thereby good commits, good change
 documentation, easier code review, faster merging) is an art and a
 habit. Just use your best judgement and we'll check and polish
 as part of CI and code review. Examples will be added here in due course.
 Related:
 - <https://groups.google.com/g/hledger/c/t14xwHQMJqU/m/9frARXIdAAAJ>
 - <https://conventionalcommits.org>
 - <https://git.savannah.gnu.org/cgit/emacs.git/plain/CONTRIBUTE> -> Commit messages
 ## Pull requests
 Most contributed hledger code (and some of the project maintainer's code)
 is submitted and reviewed via Github pull requests.
 Here are some tips for contributing PRs to hledger.
 ### Code review is important
 We aim to improve and sustain hledger's quality and maintainability over the long term.
 Many PRs, especially small ones, and even some big ones, can be merged quickly. 
 We love merging good PRs quickly.
 Some bigger or more risky PRs can require substantial review, discussion, changes, or re-submission. 
 Sometimes this is a bigger task than the coding.
 Much valuable design, quality control, and knowledge sharing happens at this time. 
 Some PRs get rejected, but their discussion and exploration can still be a useful contribution.
 We very much want to avoid wasted work, but it occasionally happens. 
 Our process is evolving and imperfect.
 All of this is normal.
 We hope you'll see it as a golden opportunity to collaborate with experts,
 share and receive knowledge, refine your design/documentation/code,
 and practice real-world development and communication skills.
 Patience and persistence pays.
 ### The pull request
 A PR should have a clear purpose, documented in its description. Mention any #ISSUENOs addressed.
 Don't tackle too much at once. 
 Smaller/more focussed PRs can be reviewed quicker and accepted (or rejected) quicker.
 Consider showing a draft of documentation first (more on this below).
 ### The commit(s)
 Commits should be easy to review.
 Ideally each commit is complete, and has a single clear purpose,
 which should be documented in the summary (and long description, if needed).
 \#ISSUENOs can be mentioned in summary/description too when appropriate.
 Within the above constraint, fewer, larger commits are preferred.
 Keep in mind that commit messages are valuable documentation 
 for future developers and troubleshooters. 
 They are also the starting point for package changelogs and hledger release notes.
 High-quality commit messages makes the release process quicker, and the resulting docs better. 
 User-impacting commits should mention the user-visible changes, 
 and be described in user-relevant language.
 Library-user-impacting commits, eg API changes, ideally will also
 be called out, and can described in more technical language.
 Commits affecting hledger internals are less important, 
 but you may notice some adhoc conventions if you browse the history.
 In particular, you can optionally prefix the summary with short component codes (cf [Issues](#issues))
 to facilitate history reading and changelog/release note production.
 Rewrite and force-push your commits freely (rebase -i, push -f) to clean them up. 
 Unless we decide to squash the PR into one commit, 
 your commits will become part of hledger's history "for all time", 
 so think about future developers trying to understand them, git bisect, etc.   
 Rebase your commits against latest master for easiest review. Especially if they start to conflict.
 We like to use some conventions in commit messages when it makes sense. These aren't mandatory, but appreciated:
 - prepend a  [topic](#topics) prefix, eg `cli: ` or `journal: `, for clarity and to help with changelog production
 - prepend a semicolon (`;`) to indicate commits that need not be mentioned in changelogs/release notes (as in the Emacs project)
 - append a final `[ci skip]` line to indicate commits that need not trigger a CI build, to reduce carbon emissions from Travis.
 ### The docs
 PRs should include appropriate updates to reference documentation, unless otherwise agreed.
 Typically this means the manual source files (hledger*/hledger*.m4.md).
 It can also involve
 command line option names and descriptions,
 other --help output,
 hledger's commands list,
 hledger-ui's help dialog,
 hledger-web's help dialog,
 etc.
 Sometimes it means the developer docs, at least the ones in the main repo (READMEs).
 Reviewers can understand your PR more efficiently once proposed doc changes are provided, and may postpone it otherwise.
 We are happy to help with the docs if needed - just ask.
 Updating rendered manuals (hledger.{1,info,txt,md,html}) is not required, and probably best avoided to reduce conflicts.
 Updating other docs such as tutorials, how-tos, examples, or screenshots is not required,
 though it's welcome (may be in a different repo).
 ### Documentation first
 hledger follows documentation-driven design.
 It is in fact highly effective, and highly recommended,
 to write the new docs (help text/user manual/haddocks/developer README..) before writing any code.
 You can share a rough draft on IRC, on the mail list, in an issue comment,
 or in a "WIP" PR starting with just the proposed docs commit.
 This is often the quickest road to getting something merged into hledger.
 hledger's many parts interact in surprisingly complex ways.
 The documentation-driven working style lets us discuss, clarify and reach a good-enough consensus economically,
 after which coding/review/acceptance can go quicker.
 <!--
 changes can impact past and future users,
 ease of contribution,
 long-term maintenance costs,
 product architecture,
 compatibility with the larger plain text accounting ecosystem,
 etc.
 -->
 ### Related ideas
 [Neil Mitchell’s Blog - The One PR Per Day Rule](https://neilmitchell.blogspot.com/2019/06/the-one-pr-per-day-rule.html)
 ## Tests
 About testing in the hledger project, as of 201809.
@ -1261,114 +1060,6 @@ Project documentation lives in a number of places:
 - `doc/notes.org` has some old developer notes
 - developer reports (profiles, benchmarks, coverage..) in doc/profs, sometimes published at hledger.org/profs
 How to prepare changelogs & release notes
 Changelogs:
 - ./Shake changelogs
 - edit the new changelog items (identify, filter, move to correct changelog, deduplicate, rewrite, sort/group)
 Release notes:
 - add a new TOC entry and section in site/release-notes.md
 - copy/rewrite/summarise package changelogs 
 - note any other items of interest
 - list release contributors
 - write release summary
 ## Issues
 The hledger project\'s issue tracker is on github. It contains:
 -   BUG issues - failures in some part of the hledger project (the main
    hledger packages, docs, website..)
 -   WISH issues - feature proposals, enhancement requests
 -   uncategorised issues - we don\'t know what these are yet
 -   pull requests - proposed changes to code and docs
 Issues are also labelled according to their [topics](#topics), for organisation.
 Some loose conventions:
 - In bug titles, mention the hledger version in which the bug first appeared 
  (and avoid mentioning version numbers otherwise).
  This allows searches like
  [new issues in 1.22](https://github.com/simonmichael/hledger/issues?q=in%3Atitle+1.22+)
  and
  [regressions in 1.22](https://github.com/simonmichael/hledger/issues?q=in%3Atitle+1.22+label%3Aregression%21)
 ### Issue Urls
 -   <http://bugs.hledger.org> - show open BUG issues
 -   <http://wishes.hledger.org> - show open WISH issues
 -   <http://issues.hledger.org> - show all issues, open or closed
 -   <http://prs.hledger.org> - show open pull requests
 -   <http://bugs.hledger.org/new> - create a new issue
 ### Labels
 Labels are used to categorise:
 -   the issue\'s type: \"A BUG\" or \"A WISH\", in shades of red (The A
    makes it appear as first label)
 -   relevant subsystems/topics, in light blue. More about this below.
 -   relevant platforms, in light purple
 -   resolution if not fixed:
    \"closed:cant-reproduce/duplicate/invalid/wont-fix\", in dark grey
 -   \"bounty\", in bright yellow: issues with bountysource funding
 -   \"easy?\", in dim yellow: issues which are probably relatively easy
    to fix
 -   \"imported\" etc., in white: miscellaneous information
 ### Topics
 Short topic names, corresponding to hledger commands, input formats, output formats and other common themes,
 are used to organise things in the hledger project. In particular,
 - They are used as space saving descriptive prefixes for [commit messages](#commit-messages)
 - They can be used as prefixes for issue/PR titles
 - Issues and PRs are labelled with them (the light blue labels).
 A more or less complete list can be seen at [open issues](#open-issues)
 or in the issue tracker's labels list.
 ### Custodians
 If you are interested in helping with a particular component for a
 while, please add yourself as a custodian in Open Issues table above.
 A custodian\'s job is to help manage the issues, rally the troops, and
 drive the open issue count towards zero. The more custodians, the
 better! By dividing up the work this way, we can scale and make forward
 progress.
 ### Milestones and Projects
 Milestones are used a little bit to plan releases. In 2017 we
 experimented with projects, but in 2018 milestones are in favour again..
 ### Estimates
 You might see some experiments in estimate tracking, where some issue
 names might have a suffix noting estimated and spent time. Basic format:
 \[ESTIMATEDTOTALTASKTIME\|TIMESPENTSOFAR\]. Examples: \`\`\` \[2\] two
 hours estimated, no time spent \[..\] half an hour estimated (a dot is
 \~a quarter hour, as in timedot format) \[1d\] one day estimated (a day
 is \~4 hours) \[1w\] one week estimated (a week is \~5 days or \~20
 hours) \[3\|2\] three hours estimated, about two hours spent so far
 \[1\|1w\|2d\] first estimate one hour, second estimate one week, about
 two days spent so far \`\`\` Estimates are always for the total time
 cost (not time remaining). Estimates are not usually changed, a new
 estimate is added instead. Numbers are very approximate, but better than
 nothing.
 ### Trello
 The [trello board](http://trello.hledger.org) (trello.hledger.org) is an
 old collection of wishlist items. This should probably be considered
 deprecated.
 ## Funding
 My vision for the hledger project has always been for it to be "accountable" and "self-sustaining", possibly through new forms of incentivisation. 
@ -1418,230 +1109,19 @@ and help us to experiment.
 You are encouraged to claim your bounties,
 though you can also choose to transfer them to a new issue of your choice.
 ## Commit messages
 See [COMMITS](COMMITS.html).
 ## Issues
 See [ISSUES](ISSUES.html).
 ## Pull Requests
 See [PULLREQUESTS](PULLREQUESTS.html).
 ## Developer workflows
-### Get developer tools
+See [WORKFLOWS](WORKFLOWS.html).
 Ensure [`stack`](https://haskell-lang.org/get-started) is installed
 (or if you’re a [cabal](https://www.haskell.org/cabal/) expert, feel free to use that.)
 Ensure [`git`](https://git-scm.com) is installed. On Windows, it comes with stack.
 Here are some useful optional tools:
 - [GNU Make](https://www.gnu.org/software/make): to use the convenient [Make rules](#make).
 - [`entr`](https://www.entrproject.org/) runs arbitrary commands when files change.
 - [`ghcid`](https://hackage.haskell.org/package/ghcid) gives real-time GHC feedback as you make code changes.
 - [`shelltestrunner`](https://hackage.haskell.org/package/shelltestrunner) runs hledger's functional tests.
 - [`quickbench`](https://hackage.haskell.org/package/quickbench) measures and reports time taken by commands.
 - [`hasktags`](https://hackage.haskell.org/package/hasktags) generates tag files for quick code navigation in editors like Emacs and vim.
 - For browsing and editing Haskell code, popular tools include: Emacs, Vim, IDEA, VS Code, Atom..
 Eg:
    stack install ghcid shelltestrunner hasktags
    git clone https://github.com/simonmichael/quickbench; cd quickbench; stack install  # must run in source dir
 ### Get the code
    git clone https://github.com/simonmichael/hledger
    cd hledger
 ### Review code
 - review and discuss new [pull requests](http://prs.hledger.org) and commits on github
 - build hledger and test the latest changes in your own repo
 - read the existing [code docs and source](#quick-links)
 - send feedback or discuss via [IRC or mail list](index.html#helpfeedback)
 ### Build in place
 See also https://hledger.org/download.html#c.-build-the-development-version .
    stack build    # hledger hledger-ui ...
 This fetches the required GHC version and haskell dependencies from the default stackage snapshot (configured in `stack.yaml`), 
 then builds all hledger packages.
 This can take a while! To save time, you can build individual packages, eg just the CLI and TUI.
 Note stack does not fetch C libraries such as curses or terminfo, which you might need to install yourself, using your system's package manager.
 In case of trouble, see [download](/download.html#link-errors).
 If you want to use an older snapshot/GHC for some reason, specify one of the older stack-*.yaml files:
    stack --stack-yaml stack8.2.yaml build
 ### Run in place
    stack exec -- hledger     # ARGS...
    stack exec -- hledger-ui  # ARGS...
    stack exec -- which hledger
 ### Build and install
 This builds and also copies the hledger executables to `~/.local/bin` or the Windows equivalent
 (which you should  [add to your `$PATH`](/download.html#b)).
    stack install    # hledger hledger-ui ...
 ### Run package tests
 Runs any HUnit/doctest/easytest tests defined by each hledger package.
    stack test    # hledger ...
 ### Run package benchmarks
 Runs any performance reports defined by each hledger package.
    stack bench    # hledger ...
 ### Run quickbench benchmarks
 Times the end-user commands in `bench.sh` using quickbench.
    make bench
 ### Run functional tests
 Runs the shelltestrunner tests defined in hledger/test/, which test the hledger CLI.
    make functest
 ### Run haddock tests
 Checks for anything that would break haddock doc generation.
    make haddocktest
 Checks for the unit-tests embedded in documentation.
    make doctest
 ### Simulate Travis tests
 Locally runs tests similar to what we run on Travis CI.
    make travistest
 ### Test with all supported GHC versions/stackage snapshots
    make allsnapshotstest
 ### Use GHCI
 GHCI is GHC's [REPL](https://en.wikipedia.org/wiki/Read%E2%80%93eval%E2%80%93print_loop), 
 useful for exploring and calling code interactively.
 If you try to run GHCI (or things based on it, like ghcid)
 right after cloning the hledger repo, you might see an error about CPP macros, eg like
 [on #961](https://github.com/simonmichael/hledger/issues/961#issuecomment-459283412).
 To fix this, build the hledger packages once, eg `stack build`.
 (Or `stack build hledger` might be enough, depending what you are doing.)
 #### Get a GHCI prompt for hledger-lib:
    cd hledger-lib; stack ghci hledger-lib
 Changing into the package directory isn't actually needed, but it
 enables a custom .ghci script which sets a more useful short prompt.
 #### Get a GHCI prompt for hledger:
    cd hledger; stack ghci hledger
 #### Get a GHCI prompt for hledger-ui:
    cd hledger-ui; stack ghci hledger-ui
 #### Get a GHCI prompt for hledger-web:
    cd hledger-web; stack ghci hledger-web
 hledger-web also needs to find some things in its current directory (like the static/ directory).
 This normally just works, if not please [send details](https://github.com/simonmichael/hledger/issues/274).
 ### Add a test
 - identify what to test
 - choose the test type: unit ? functional ? benchmark ?
 - currently expected to pass or fail ?
 - figure out where it goes
 - write test, verify expected result
 - get it committed
 ### Fix a bug or add a feature
 - research, discuss, validate the issue/feature on irc/list/bug tracker
 - look for related tests, run the tests and check they are passing
 - add a test ?
 - develop a patch
 - include any related issue numbers in the patch name, eg: "fix for blah blah (#NNN)"
 - get it committed
 ### Get your changes accepted
 Follow the usual github workflow:
 - fork the main hledger repo on github,
 - git clone it to your local machine,
 - git commit, after (?) pulling and merging the latest upstream changes
 - git push back to github,
 - open a pull request on github,
 - follow up on any discussion there.
 If you're new to this process, [help.github.com](https://help.github.com) may be useful.
 ### Add yourself to the contributor list
 - after getting something into the master branch, read and sign the [contributor list & agreement](https://hledger.org/contributors.html). Or, [ask](index.html#help-feedback) to be added.
 - give yourself a high five!
 ### Work on docs
 Most docs tasks are handled by [Shake](#shake). 
 #### List Shake rules:
    ./Shake
 #### Generate man/info/txt manuals (in hledger*/) and embed in hledger executables:
    ./Shake manuals
    stack build
 #### Generate html manuals and the hledger website (in site/_site/):
    ./Shake website
 #### To remove all files generated by Shake:
    ./Shake Clean
 ### Use ghcid for watching GHC/GHCI
 [ghcid](https://hackage.haskell.org/package/ghcid) is the most reliable and fastest way to see GHC's feedback, and optionally run tests or a GHCI command, as you edit. We run it via make, for convenience and to watch multiple packages rather than just one. Run `make help-ghcid` to list related rules.
 #### Watch for compile errors in hledger-lib and hledger:
    make ghcid
 #### Watch compile errors and the output of some hledger command:
    ghcid -c 'make ghci' -T ':main -f a.j bal --budget -N'
 ### Use --file-watch for watching stack
 stack's --file-watch flag will re-run build/test/bench when source files or package.yaml/cabal files change. Eg:
    stack test hledger --file-watch
 If you find that adding --fast makes this any faster, please update this.
 ### Use entr for watching arbitrary commands
 [entr](https://entrproject.org/) is the most robust cross-platform tool for watching files and running a command when they change. Note its first argument must be an executable program, to run a shell command or multiple commands use `bash -c "..."`.
 #### Rerun a single functional test as you change it:
    ls hledger/test/budget/budget.test | entr bash -c 'clear; COLUMNS=80 stack exec -- shelltest --execdir hledger/test/budget/budget.test -i12'
--- a/ISSUES.md
+++ b/ISSUES.md
@ -0,0 +1,96 @@
 # Issues
 <div class=pagetoc>
 <!-- toc -->
 </div>
 The hledger project\'s issue tracker is on github. It contains:
 -   BUG issues - failures in some part of the hledger project (the main
    hledger packages, docs, website..)
 -   WISH issues - feature proposals, enhancement requests
 -   uncategorised issues - we don\'t know what these are yet
 -   pull requests - proposed changes to code and docs
 Issues are also labelled according to their [topics](#topics), for organisation.
 Some loose conventions:
 - In bug titles, mention the hledger version in which the bug first appeared 
  (and avoid mentioning version numbers otherwise).
  This allows searches like
  [new issues in 1.22](https://github.com/simonmichael/hledger/issues?q=in%3Atitle+1.22+)
  and
  [regressions in 1.22](https://github.com/simonmichael/hledger/issues?q=in%3Atitle+1.22+label%3Aregression%21)
 ## Issue Urls
 -   <http://bugs.hledger.org> - show open BUG issues
 -   <http://wishes.hledger.org> - show open WISH issues
 -   <http://issues.hledger.org> - show all issues, open or closed
 -   <http://prs.hledger.org> - show open pull requests
 -   <http://bugs.hledger.org/new> - create a new issue
 ## Labels
 Labels are used to categorise:
 -   the issue\'s type: \"A BUG\" or \"A WISH\", in shades of red (The A
    makes it appear as first label)
 -   relevant subsystems/topics, in light blue. More about this below.
 -   relevant platforms, in light purple
 -   resolution if not fixed:
    \"closed:cant-reproduce/duplicate/invalid/wont-fix\", in dark grey
 -   \"bounty\", in bright yellow: issues with bountysource funding
 -   \"easy?\", in dim yellow: issues which are probably relatively easy
    to fix
 -   \"imported\" etc., in white: miscellaneous information
 ## Topics
 Short topic names, corresponding to hledger commands, input formats, output formats and other common themes,
 are used to organise things in the hledger project. In particular,
 - They are used as space saving descriptive prefixes for [commit messages](#commit-messages)
 - They can be used as prefixes for issue/PR titles
 - Issues and PRs are labelled with them (the light blue labels).
 A more or less complete list can be seen at [open issues](#open-issues)
 or in the issue tracker's labels list.
 ## Custodians
 If you are interested in helping with a particular component for a
 while, please add yourself as a custodian in Open Issues table above.
 A custodian\'s job is to help manage the issues, rally the troops, and
 drive the open issue count towards zero. The more custodians, the
 better! By dividing up the work this way, we can scale and make forward
 progress.
 ## Milestones and Projects
 Milestones are used a little bit to plan releases. In 2017 we
 experimented with projects, but in 2018 milestones are in favour again..
 ## Estimates
 You might see some experiments in estimate tracking, where some issue
 names might have a suffix noting estimated and spent time. Basic format:
 \[ESTIMATEDTOTALTASKTIME\|TIMESPENTSOFAR\]. Examples: \`\`\` \[2\] two
 hours estimated, no time spent \[..\] half an hour estimated (a dot is
 \~a quarter hour, as in timedot format) \[1d\] one day estimated (a day
 is \~4 hours) \[1w\] one week estimated (a week is \~5 days or \~20
 hours) \[3\|2\] three hours estimated, about two hours spent so far
 \[1\|1w\|2d\] first estimate one hour, second estimate one week, about
 two days spent so far \`\`\` Estimates are always for the total time
 cost (not time remaining). Estimates are not usually changed, a new
 estimate is added instead. Numbers are very approximate, but better than
 nothing.
 ## Trello
 The [trello board](http://trello.hledger.org) (trello.hledger.org) is an
 old collection of wishlist items. This should probably be considered
 deprecated.
--- a/PULLREQUESTS.md
+++ b/PULLREQUESTS.md
@ -0,0 +1,117 @@
 # Pull requests
 Most contributed hledger code (and some of the project maintainer's code)
 is submitted and reviewed via Github pull requests.
 Here are some tips for contributing PRs to hledger.
 ## Code review is important
 We aim to improve and sustain hledger's quality and maintainability over the long term.
 Many PRs, especially small ones, and even some big ones, can be merged quickly. 
 We love merging good PRs quickly.
 Some bigger or more risky PRs can require substantial review, discussion, changes, or re-submission. 
 Sometimes this is a bigger task than the coding.
 Much valuable design, quality control, and knowledge sharing happens at this time. 
 Some PRs get rejected, but their discussion and exploration can still be a useful contribution.
 We very much want to avoid wasted work, but it occasionally happens. 
 Our process is evolving and imperfect.
 All of this is normal.
 We hope you'll see it as a golden opportunity to collaborate with experts,
 share and receive knowledge, refine your design/documentation/code,
 and practice real-world development and communication skills.
 Patience and persistence pays.
 ## The pull request
 A PR should have a clear purpose, documented in its description. Mention any #ISSUENOs addressed.
 Don't tackle too much at once. 
 Smaller/more focussed PRs can be reviewed quicker and accepted (or rejected) quicker.
 Consider showing a draft of documentation first (more on this below).
 ## The commit(s)
 Commits should be easy to review.
 Ideally each commit is complete, and has a single clear purpose,
 which should be documented in the summary (and long description, if needed).
 \#ISSUENOs can be mentioned in summary/description too when appropriate.
 Within the above constraint, fewer, larger commits are preferred.
 Keep in mind that commit messages are valuable documentation 
 for future developers and troubleshooters. 
 They are also the starting point for package changelogs and hledger release notes.
 High-quality commit messages makes the release process quicker, and the resulting docs better. 
 User-impacting commits should mention the user-visible changes, 
 and be described in user-relevant language.
 Library-user-impacting commits, eg API changes, ideally will also
 be called out, and can described in more technical language.
 Commits affecting hledger internals are less important, 
 but you may notice some adhoc conventions if you browse the history.
 In particular, you can optionally prefix the summary with short component codes (cf [Issues](#issues))
 to facilitate history reading and changelog/release note production.
 Rewrite and force-push your commits freely (rebase -i, push -f) to clean them up. 
 Unless we decide to squash the PR into one commit, 
 your commits will become part of hledger's history "for all time", 
 so think about future developers trying to understand them, git bisect, etc.   
 Rebase your commits against latest master for easiest review. Especially if they start to conflict.
 We like to use some conventions in commit messages when it makes sense. These aren't mandatory, but appreciated:
 - prepend a  [topic](#topics) prefix, eg `cli: ` or `journal: `, for clarity and to help with changelog production
 - prepend a semicolon (`;`) to indicate commits that need not be mentioned in changelogs/release notes (as in the Emacs project)
 - append a final `[ci skip]` line to indicate commits that need not trigger a CI build, to reduce carbon emissions from Travis.
 ## The docs
 PRs should include appropriate updates to reference documentation, unless otherwise agreed.
 Typically this means the manual source files (hledger*/hledger*.m4.md).
 It can also involve
 command line option names and descriptions,
 other --help output,
 hledger's commands list,
 hledger-ui's help dialog,
 hledger-web's help dialog,
 etc.
 Sometimes it means the developer docs, at least the ones in the main repo (READMEs).
 Reviewers can understand your PR more efficiently once proposed doc changes are provided, and may postpone it otherwise.
 We are happy to help with the docs if needed - just ask.
 Updating rendered manuals (hledger.{1,info,txt,md,html}) is not required, and probably best avoided to reduce conflicts.
 Updating other docs such as tutorials, how-tos, examples, or screenshots is not required,
 though it's welcome (may be in a different repo).
 ## Documentation first
 hledger follows documentation-driven design.
 It is in fact highly effective, and highly recommended,
 to write the new docs (help text/user manual/haddocks/developer README..) before writing any code.
 You can share a rough draft on IRC, on the mail list, in an issue comment,
 or in a "WIP" PR starting with just the proposed docs commit.
 This is often the quickest road to getting something merged into hledger.
 hledger's many parts interact in surprisingly complex ways.
 The documentation-driven working style lets us discuss, clarify and reach a good-enough consensus economically,
 after which coding/review/acceptance can go quicker.
 <!--
 changes can impact past and future users,
 ease of contribution,
 long-term maintenance costs,
 product architecture,
 compatibility with the larger plain text accounting ecosystem,
 etc.
 -->
 ## Related ideas
 [Neil Mitchell’s Blog - The One PR Per Day Rule](https://neilmitchell.blogspot.com/2019/06/the-one-pr-per-day-rule.html)
--- a/WORKFLOWS.md
+++ b/WORKFLOWS.md
@ -0,0 +1,231 @@
 # Developer workflows
 <div class=pagetoc>
 <!-- toc -->
 </div>
 ## Get developer tools
 Ensure [`stack`](https://haskell-lang.org/get-started) is installed
 (or if you’re a [cabal](https://www.haskell.org/cabal/) expert, feel free to use that.)
 Ensure [`git`](https://git-scm.com) is installed. On Windows, it comes with stack.
 Here are some useful optional tools:
 - [GNU Make](https://www.gnu.org/software/make): to use the convenient [Make rules](#make).
 - [`entr`](https://www.entrproject.org/) runs arbitrary commands when files change.
 - [`ghcid`](https://hackage.haskell.org/package/ghcid) gives real-time GHC feedback as you make code changes.
 - [`shelltestrunner`](https://hackage.haskell.org/package/shelltestrunner) runs hledger's functional tests.
 - [`quickbench`](https://hackage.haskell.org/package/quickbench) measures and reports time taken by commands.
 - [`hasktags`](https://hackage.haskell.org/package/hasktags) generates tag files for quick code navigation in editors like Emacs and vim.
 - For browsing and editing Haskell code, popular tools include: Emacs, Vim, IDEA, VS Code, Atom..
 Eg:
    stack install ghcid shelltestrunner hasktags
    git clone https://github.com/simonmichael/quickbench; cd quickbench; stack install  # must run in source dir
 ## Get the code
    git clone https://github.com/simonmichael/hledger
    cd hledger
 ## Review code
 - review and discuss new [pull requests](http://prs.hledger.org) and commits on github
 - build hledger and test the latest changes in your own repo
 - read the existing [code docs and source](#quick-links)
 - send feedback or discuss via [IRC or mail list](index.html#helpfeedback)
 ## Build in place
 See also https://hledger.org/download.html#c.-build-the-development-version .
    stack build    # hledger hledger-ui ...
 This fetches the required GHC version and haskell dependencies from the default stackage snapshot (configured in `stack.yaml`), 
 then builds all hledger packages.
 This can take a while! To save time, you can build individual packages, eg just the CLI and TUI.
 Note stack does not fetch C libraries such as curses or terminfo, which you might need to install yourself, using your system's package manager.
 In case of trouble, see [download](/download.html#link-errors).
 If you want to use an older snapshot/GHC for some reason, specify one of the older stack-*.yaml files:
    stack --stack-yaml stack8.2.yaml build
 ## Run in place
    stack exec -- hledger     # ARGS...
    stack exec -- hledger-ui  # ARGS...
    stack exec -- which hledger
 ## Build and install
 This builds and also copies the hledger executables to `~/.local/bin` or the Windows equivalent
 (which you should  [add to your `$PATH`](/download.html#b)).
    stack install    # hledger hledger-ui ...
 ## Run package tests
 Runs any HUnit/doctest/easytest tests defined by each hledger package.
    stack test    # hledger ...
 ## Run package benchmarks
 Runs any performance reports defined by each hledger package.
    stack bench    # hledger ...
 ## Run quickbench benchmarks
 Times the end-user commands in `bench.sh` using quickbench.
    make bench
 ## Run functional tests
 Runs the shelltestrunner tests defined in hledger/test/, which test the hledger CLI.
    make functest
 ## Run haddock tests
 Checks for anything that would break haddock doc generation.
    make haddocktest
 Checks for the unit-tests embedded in documentation.
    make doctest
 ## Simulate Travis tests
 Locally runs tests similar to what we run on Travis CI.
    make travistest
 ## Test with all supported GHC versions/stackage snapshots
    make allsnapshotstest
 ## Use GHCI
 GHCI is GHC's [REPL](https://en.wikipedia.org/wiki/Read%E2%80%93eval%E2%80%93print_loop), 
 useful for exploring and calling code interactively.
 If you try to run GHCI (or things based on it, like ghcid)
 right after cloning the hledger repo, you might see an error about CPP macros, eg like
 [on #961](https://github.com/simonmichael/hledger/issues/961#issuecomment-459283412).
 To fix this, build the hledger packages once, eg `stack build`.
 (Or `stack build hledger` might be enough, depending what you are doing.)
 ### Get a GHCI prompt for hledger-lib:
    cd hledger-lib; stack ghci hledger-lib
 Changing into the package directory isn't actually needed, but it
 enables a custom .ghci script which sets a more useful short prompt.
 ### Get a GHCI prompt for hledger:
    cd hledger; stack ghci hledger
 ### Get a GHCI prompt for hledger-ui:
    cd hledger-ui; stack ghci hledger-ui
 ### Get a GHCI prompt for hledger-web:
    cd hledger-web; stack ghci hledger-web
 hledger-web also needs to find some things in its current directory (like the static/ directory).
 This normally just works, if not please [send details](https://github.com/simonmichael/hledger/issues/274).
 ## Add a test
 - identify what to test
 - choose the test type: unit ? functional ? benchmark ?
 - currently expected to pass or fail ?
 - figure out where it goes
 - write test, verify expected result
 - get it committed
 ## Fix a bug or add a feature
 - research, discuss, validate the issue/feature on irc/list/bug tracker
 - look for related tests, run the tests and check they are passing
 - add a test ?
 - develop a patch
 - include any related issue numbers in the patch name, eg: "fix for blah blah (#NNN)"
 - get it committed
 ## Get your changes accepted
 Follow the usual github workflow:
 - fork the main hledger repo on github,
 - git clone it to your local machine,
 - git commit, after (?) pulling and merging the latest upstream changes
 - git push back to github,
 - open a pull request on github,
 - follow up on any discussion there.
 If you're new to this process, [help.github.com](https://help.github.com) may be useful.
 ## Add yourself to the contributor list
 - after getting something into the master branch, read and sign the [contributor list & agreement](https://hledger.org/contributors.html). Or, [ask](index.html#help-feedback) to be added.
 - give yourself a high five!
 ## Work on docs
 Most docs tasks are handled by [Shake](#shake). 
 ### List Shake rules:
    ./Shake
 ### Generate man/info/txt manuals (in hledger*/) and embed in hledger executables:
    ./Shake manuals
    stack build
 ### Generate html manuals and the hledger website (in site/_site/):
    ./Shake website
 ### To remove all files generated by Shake:
    ./Shake Clean
 ## Use ghcid for watching GHC/GHCI
 [ghcid](https://hackage.haskell.org/package/ghcid) is the most reliable and fastest way to see GHC's feedback, and optionally run tests or a GHCI command, as you edit. We run it via make, for convenience and to watch multiple packages rather than just one. Run `make help-ghcid` to list related rules.
 ### Watch for compile errors in hledger-lib and hledger:
    make ghcid
 ### Watch compile errors and the output of some hledger command:
    ghcid -c 'make ghci' -T ':main -f a.j bal --budget -N'
 ## Use --file-watch for watching stack
 stack's --file-watch flag will re-run build/test/bench when source files or package.yaml/cabal files change. Eg:
    stack test hledger --file-watch
 If you find that adding --fast makes this any faster, please update this.
 ## Use entr for watching arbitrary commands
 [entr](https://entrproject.org/) is the most robust cross-platform tool for watching files and running a command when they change. Note its first argument must be an executable program, to run a shell command or multiple commands use `bash -c "..."`.
 ### Rerun a single functional test as you change it:
    ls hledger/test/budget/budget.test | entr bash -c 'clear; COLUMNS=80 stack exec -- shelltest --execdir hledger/test/budget/budget.test -i12'