Extend nvim-cmp to auto-complete Jira issue keys.

Based on cmp_gh_source example by TJ DeVries.

• Using NeoVim (tested with 0.9.0 on unix-like environment)
• Access to Jira-Server REST API (not tested with Jira-Cloud)
• curl is installed and is available in $PATH
• nvim-cmp and plenary.nvim are installed (both must be loaded before setup is called)

Use plugin manager of your choice. Minimal installation example for lazy.nvim:

require('lazy').setup({
  {
    'artem-nefedov/cmp-jira-issues.nvim',
    dependencies = { -- this is not required if you already install those plugins
      'nvim-lua/plenary.nvim',
      'hrsh7th/nvim-cmp',
    },
  },
})

You must have a curl config file (default: ~/.jira-curl-config), which must contain at least the following options:

• url - value must point to your Jira server's /rest/api/2/search endpoint
• data-urlencode - value must have jql=<your issue search query>
• any other options required to authenticate to your Jira server (header, cookie, user, etc.)

The file will be passed to curl's --config option as-is.

This example shows how to get open issues assigned to you for completion:

url = "https://your-jira-server/rest/api/2/search"
data-urlencode = "jql=assignee = \"your-acc-name\" and resolution = unresolved"
user = "username:password"
# instead of "user", you can use any other option that can control authentication
# e.g. header = "Authorization: Bearer <your-personal-access-token>"

Make sure to call setup when plenary and nvim-cmp are already loaded.

Create/update ".lua" file in ~/.config/nvim/after/plugin/ with contents:

require('cmp-jira-issues').setup({})

Don't forget to add completion source (default: jira_issues) to cmp setup, e.g.:

local cmp = require('cmp')

cmp.setup({
  -- other options
  sources = {
    { name = 'jira_issues' },
    -- other sources
  },
})

Below you can see available fields (shown with values set to match defaults):

require('cmp-jira-issues').setup({
  source_name = 'jira_issues',
  get_trigger_characters = function() -- on which characters completion is triggered
    return { '[' }
  end,
  is_available = function() -- determine whether source is available for current buffer
    return vim.bo.filetype == 'gitcommit' or
        (vim.bo.filetype == 'markdown' and vim.fs.basename(vim.api.nvim_buf_get_name(0)) == 'CHANGELOG.md')
  end,
  clear_cache = function() -- set to boolean false value to disable user command creation
    vim.g.cached_jira_issues = nil
  end,
  complete_opts = {
    curl_config = '~/.jira-curl-config', -- value is passed to `:h expand()`
    fields = 'summary,description', -- what fields to fetch from jira api
    items = { -- what fields to lookup in response and how to format them
      { '[%s] ',   { { 'key' } } }, -- key only
      { '[%s] %s', { { 'key' }, { 'fields', 'summary' } } }, -- key + summary
    },
    get_cache = function(_, _)
      return vim.g.cached_jira_issues
    end,
    set_cache = function(_, _, items)
      vim.g.cached_jira_issues = items
    end,
  },
})

You can change complete_opts.items to change what values will be returned, e.g. this value will make it so only return lines with both issue keys and summary are returned:

items = {
  { '[%s] %s', { { 'key' }, { 'fields', 'summary' } } },
}

Number of %s in LHS must match total number of defined elements in RHS.

Note that RHS is a list which itself contains other lists of values that correspond to structure of returned "issue" JSON object, with arbitrary depth, e.g. { 'fields', 'summary' } means get value from issue.fields.summary.

By default, response from server is cached globally for the entire session duration, and JiraClearCache user command is provided to clear cached results.

You can change the behavior by implementing your own caching mechanics using complete_opts.get_cache, complete_opt.set_cache, and clear_cache callbacks.

• complete_opts.get_cache receives 2 arguments: self table and bufnr integer
• complete_opt.set_cache receives 3 arguments: self table, bufnr integer, and items table
• clear_cache doesn't receive anything and gets called by JiraClearCache command

Furthermore, self table has child cache table created by default.

To implement buffer local cache, use the following definitions (note that it makes clear_cache useless - set it to false to disable user command creation):

get_cache = function(self, bufnr)
  return self.cache[bufnr]
end

set_cache = function(self, bufnr, items)
  self.cache[bufnr] = items
end

To disable cache completely, pass complete_opts.get_cache function that always returns nil.

If curl doesn't return correct results, debug it by running it in local shell with the same arguments as seen in complete.lua.