////////////////////////////////////////////////////////////////////////////////
//
// Copyright 2006 - 2021, Tomas Babej, Paul Beckingham, Federico Hernandez.
//
// Permission is hereby granted, free of charge, to any person obtaining a copy
// of this software and associated documentation files (the "Software"), to deal
// in the Software without restriction, including without limitation the rights
// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
// copies of the Software, and to permit persons to whom the Software is
// furnished to do so, subject to the following conditions:
//
// The above copyright notice and this permission notice shall be included
// in all copies or substantial portions of the Software.
//
// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
// OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
// THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
// SOFTWARE.
//
// https://www.opensource.org/licenses/mit-license.php
//
////////////////////////////////////////////////////////////////////////////////

#include <cmake.h>
// cmake.h include header must come first

#include <Task.h>
#include <assert.h>
#include <stdlib.h>

#include <sstream>
#include <string>
#ifdef PRODUCT_TASKWARRIOR
#include <ctype.h>
#include <math.h>
#endif
#include <Lexer.h>

#include <algorithm>
#include <cfloat>
#ifdef PRODUCT_TASKWARRIOR
#include <Context.h>
#include <Pig.h>
#endif
#include <Datetime.h>
#include <Duration.h>
#ifdef PRODUCT_TASKWARRIOR
#include <RX.h>
#endif
#include <format.h>
#include <shared.h>
#include <util.h>

#ifdef PRODUCT_TASKWARRIOR
#include <Eval.h>
#include <Filter.h>
#include <Variant.h>
#include <dependency.h>
#include <feedback.h>

#define APPROACHING_INFINITY 1000  // Close enough.  This isn't rocket surgery.

static const float epsilon = 0.000001;
#endif

std::string Task::defaultProject = "";
std::string Task::defaultDue = "";
std::string Task::defaultScheduled = "";
bool Task::searchCaseSensitive = true;
bool Task::regex = false;
std::map<std::string, std::string> Task::attributes;

std::map<std::string, float> Task::coefficients;
float Task::urgencyProjectCoefficient = 0.0;
float Task::urgencyActiveCoefficient = 0.0;
float Task::urgencyScheduledCoefficient = 0.0;
float Task::urgencyWaitingCoefficient = 0.0;
float Task::urgencyBlockedCoefficient = 0.0;
float Task::urgencyAnnotationsCoefficient = 0.0;
float Task::urgencyTagsCoefficient = 0.0;
float Task::urgencyDueCoefficient = 0.0;
float Task::urgencyBlockingCoefficient = 0.0;
float Task::urgencyAgeCoefficient = 0.0;
float Task::urgencyAgeMax = 0.0;

std::map<std::string, std::vector<std::string>> Task::customOrder;

static const std::string dummy("");

////////////////////////////////////////////////////////////////////////////////
// The uuid and id attributes must be exempt from comparison.
//
// This performs two tests which are sufficient and necessary for Task
// object equality (neglecting uuid and id):
//     - The attribute set sizes are the same
//     - For each attribute in the first set, there exists a same
//       attribute with a same value in the second set
//
// These two conditions are necessary. They are also sufficient, since there
// can be no extra data attribute in the second set, due to the same attribute
// set sizes.
bool Task::operator==(const Task& other) {
  if (data.size() != other.data.size()) return false;

  for (const auto& i : data)
    if (i.first != "uuid" && i.second != other.get(i.first)) return false;

  return true;
}

////////////////////////////////////////////////////////////////////////////////
bool Task::operator!=(const Task& other) { return !(*this == other); }

////////////////////////////////////////////////////////////////////////////////
Task::Task(const std::string& input) {
  id = 0;
  urgency_value = 0.0;
  recalc_urgency = true;
  is_blocked = false;
  is_blocking = false;
  annotation_count = 0;

  parse(input);
}

////////////////////////////////////////////////////////////////////////////////
Task::Task(const json::object* obj) {
  id = 0;
  urgency_value = 0.0;
  recalc_urgency = true;
  is_blocked = false;
  is_blocking = false;
  annotation_count = 0;

  parseJSON(obj);
}

////////////////////////////////////////////////////////////////////////////////
Task::Task(rust::Box<tc::TaskData> obj) {
  id = 0;
  urgency_value = 0.0;
  recalc_urgency = true;
  is_blocked = false;
  is_blocking = false;
  annotation_count = 0;

  parseTC(std::move(obj));
}

////////////////////////////////////////////////////////////////////////////////
Task::status Task::textToStatus(const std::string& input) {
  if (input[0] == 'p')
    return Task::pending;
  else if (input[0] == 'c')
    return Task::completed;
  else if (input[0] == 'd')
    return Task::deleted;
  else if (input[0] == 'r')
    return Task::recurring;
  // for compatibility, parse `w` as pending; Task::getStatus will
  // apply the virtual waiting status if appropriate
  else if (input[0] == 'w')
    return Task::pending;

  throw format("The status '{1}' is not valid.", input);
}

////////////////////////////////////////////////////////////////////////////////
std::string Task::statusToText(Task::status s) {
  if (s == Task::pending)
    return "pending";
  else if (s == Task::recurring)
    return "recurring";
  else if (s == Task::waiting)
    return "waiting";
  else if (s == Task::completed)
    return "completed";
  else if (s == Task::deleted)
    return "deleted";

  return "pending";
}

////////////////////////////////////////////////////////////////////////////////
// Returns a proper handle to the task. Tasks should not be referenced by UUIDs
// as long as they have non-zero ID.
const std::string Task::identifier(bool shortened /* = false */) const {
  if (id != 0)
    return format(id);
  else if (shortened)
    return get("uuid").substr(0, 8);
  else
    return get("uuid");
}

////////////////////////////////////////////////////////////////////////////////
void Task::setAsNow(const std::string& att) {
  char now[22];
  snprintf(now, 22, "%lli", (long long int)time(nullptr));
  set(att, now);

  recalc_urgency = true;
}

////////////////////////////////////////////////////////////////////////////////
bool Task::has(const std::string& name) const {
  if (data.find(name) != data.end()) return true;

  return false;
}

////////////////////////////////////////////////////////////////////////////////
std::vector<std::string> Task::all() const {
  std::vector<std::string> all;
  for (const auto& i : data) all.push_back(i.first);

  return all;
}

////////////////////////////////////////////////////////////////////////////////
const std::string Task::get(const std::string& name) const {
  auto i = data.find(name);
  if (i != data.end()) return i->second;

  return "";
}

////////////////////////////////////////////////////////////////////////////////
const std::string& Task::get_ref(const std::string& name) const {
  auto i = data.find(name);
  if (i != data.end()) return i->second;

  return dummy;
}

////////////////////////////////////////////////////////////////////////////////
int Task::get_int(const std::string& name) const {
  auto i = data.find(name);
  if (i != data.end()) return strtol(i->second.c_str(), nullptr, 10);

  return 0;
}

////////////////////////////////////////////////////////////////////////////////
unsigned long Task::get_ulong(const std::string& name) const {
  auto i = data.find(name);
  if (i != data.end()) return strtoul(i->second.c_str(), nullptr, 10);

  return 0;
}

////////////////////////////////////////////////////////////////////////////////
float Task::get_float(const std::string& name) const {
  auto i = data.find(name);
  if (i != data.end()) return strtof(i->second.c_str(), nullptr);

  return 0.0;
}

////////////////////////////////////////////////////////////////////////////////
time_t Task::get_date(const std::string& name) const {
  auto i = data.find(name);
  if (i != data.end()) return (time_t)strtoul(i->second.c_str(), nullptr, 10);

  return 0;
}

////////////////////////////////////////////////////////////////////////////////
void Task::set(const std::string& name, const std::string& value) {
  data[name] = value;

  if (isAnnotationAttr(name)) ++annotation_count;

  recalc_urgency = true;
}

////////////////////////////////////////////////////////////////////////////////
void Task::set(const std::string& name, long long value) {
  data[name] = format(value);

  recalc_urgency = true;
}

////////////////////////////////////////////////////////////////////////////////
void Task::remove(const std::string& name) {
  if (data.erase(name)) recalc_urgency = true;

  if (isAnnotationAttr(name)) --annotation_count;
}

////////////////////////////////////////////////////////////////////////////////
Task::status Task::getStatus() const {
  if (!has("status")) return Task::pending;

  auto status = textToStatus(get("status"));

  // Implement the "virtual" Task::waiting status, which is not stored on-disk
  // but is defined as a pending task with a `wait` attribute in the future.
  // This is workaround for 2.6.0, remove in 3.0.0.
  if (status == Task::pending && is_waiting()) {
    return Task::waiting;
  }

  return status;
}

////////////////////////////////////////////////////////////////////////////////
void Task::setStatus(Task::status status) {
  // the 'waiting' status is a virtual version of 'pending', so translate
  // that back to 'pending' here
  if (status == Task::waiting) status = Task::pending;

  set("status", statusToText(status));

  recalc_urgency = true;
}

#ifdef PRODUCT_TASKWARRIOR
////////////////////////////////////////////////////////////////////////////////
// Determines status of a date attribute.
Task::dateState Task::getDateState(const std::string& name) const {
  time_t value = get_date(name);
  if (value > 0) {
    Datetime reference(value);
    Datetime now;
    Datetime today("today");

    if (reference < today) return dateBeforeToday;

    if (reference.sameDay(now)) {
      if (reference < now)
        return dateEarlierToday;
      else
        return dateLaterToday;
    }

    int imminentperiod = Context::getContext().config.getInteger("due");
    if (imminentperiod == 0) return dateAfterToday;

    Datetime imminentDay = today + imminentperiod * 86400;
    if (reference < imminentDay) return dateAfterToday;
  }

  return dateNotDue;
}

////////////////////////////////////////////////////////////////////////////////
// An empty task is typically a "dummy", such as in DOM evaluation, which may or
// may not occur in the context of a task.
bool Task::is_empty() const { return data.size() == 0; }

////////////////////////////////////////////////////////////////////////////////
// Ready means pending, not blocked and either not scheduled or scheduled before
// now.
bool Task::is_ready() const {
  return getStatus() == Task::pending && !is_blocked &&
         (!has("scheduled") || Datetime("now").operator>(get_date("scheduled")));
}

////////////////////////////////////////////////////////////////////////////////
bool Task::is_due() const {
  if (has("due")) {
    Task::status status = getStatus();

    if (status != Task::completed && status != Task::deleted) {
      Task::dateState state = getDateState("due");
      if (state == dateAfterToday || state == dateEarlierToday || state == dateLaterToday)
        return true;
    }
  }

  return false;
}

////////////////////////////////////////////////////////////////////////////////
bool Task::is_dueyesterday() const {
  if (has("due")) {
    Task::status status = getStatus();

    if (status != Task::completed && status != Task::deleted) {
      if (Datetime("yesterday").sameDay(get_date("due"))) return true;
    }
  }

  return false;
}

////////////////////////////////////////////////////////////////////////////////
bool Task::is_duetoday() const {
  if (has("due")) {
    Task::status status = getStatus();

    if (status != Task::completed && status != Task::deleted) {
      Task::dateState state = getDateState("due");
      if (state == dateEarlierToday || state == dateLaterToday) return true;
    }
  }

  return false;
}

////////////////////////////////////////////////////////////////////////////////
bool Task::is_duetomorrow() const {
  if (has("due")) {
    Task::status status = getStatus();

    if (status != Task::completed && status != Task::deleted) {
      if (Datetime("tomorrow").sameDay(get_date("due"))) return true;
    }
  }

  return false;
}

////////////////////////////////////////////////////////////////////////////////
bool Task::is_dueweek() const {
  if (has("due")) {
    Task::status status = getStatus();

    if (status != Task::completed && status != Task::deleted) {
      Datetime due(get_date("due"));
      if (due >= Datetime("sow") && due <= Datetime("eow")) return true;
    }
  }

  return false;
}

////////////////////////////////////////////////////////////////////////////////
bool Task::is_duemonth() const {
  if (has("due")) {
    Task::status status = getStatus();

    if (status != Task::completed && status != Task::deleted) {
      Datetime due(get_date("due"));
      if (due >= Datetime("som") && due <= Datetime("eom")) return true;
    }
  }

  return false;
}

////////////////////////////////////////////////////////////////////////////////
bool Task::is_duequarter() const {
  if (has("due")) {
    Task::status status = getStatus();

    if (status != Task::completed && status != Task::deleted) {
      Datetime due(get_date("due"));
      if (due >= Datetime("soq") && due <= Datetime("eoq")) return true;
    }
  }

  return false;
}

////////////////////////////////////////////////////////////////////////////////
bool Task::is_dueyear() const {
  if (has("due")) {
    Task::status status = getStatus();

    if (status != Task::completed && status != Task::deleted) {
      Datetime due(get_date("due"));
      if (due >= Datetime("soy") && due <= Datetime("eoy")) return true;
    }
  }

  return false;
}

////////////////////////////////////////////////////////////////////////////////
bool Task::is_udaPresent() const {
  for (auto& col : Context::getContext().columns)
    if (col.second->is_uda() && has(col.first)) return true;

  return false;
}

////////////////////////////////////////////////////////////////////////////////
bool Task::is_orphanPresent() const {
  for (auto& att : data)
    if (!isAnnotationAttr(att.first) && !isTagAttr(att.first) && !isDepAttr(att.first) &&
        Context::getContext().columns.find(att.first) == Context::getContext().columns.end())
      return true;

  return false;
}

////////////////////////////////////////////////////////////////////////////////
bool Task::is_overdue() const {
  if (has("due")) {
    Task::status status = getStatus();

    if (status != Task::completed && status != Task::deleted && status != Task::recurring) {
      Task::dateState state = getDateState("due");
      if (state == dateEarlierToday || state == dateBeforeToday) return true;
    }
  }

  return false;
}
#endif

////////////////////////////////////////////////////////////////////////////////
// Task is considered waiting if it's pending and the wait attribute is set as
// future datetime value.
// While this is not consistent with other attribute-based virtual tags, such
// as +BLOCKED, it is more backwards compatible with how +WAITING virtual tag
// behaved in the past, when waiting had a dedicated status value.
bool Task::is_waiting() const {
  if (has("wait") && get("status") == "pending") {
    Datetime now;
    Datetime wait(get_date("wait"));
    if (wait > now) return true;
  }

  return false;
}

////////////////////////////////////////////////////////////////////////////////
// Try a JSON parse.
void Task::parse(const std::string& input) {
  parseJSON(input);

  // for compatibility, include all tags in `tags` as `tag_..` attributes
  if (data.find("tags") != data.end()) {
    for (auto& tag : split(data["tags"], ',')) {
      data[tag2Attr(tag)] = "x";
    }
  }
  // ..and similarly, update `tags` to match the `tag_..` attributes
  fixTagsAttribute();

  // same for `depends` / `dep_..`
  if (data.find("depends") != data.end()) {
    for (auto& dep : split(data["depends"], ',')) {
      data[dep2Attr(dep)] = "x";
    }
  }
  fixDependsAttribute();

  recalc_urgency = true;
}

////////////////////////////////////////////////////////////////////////////////
// Note that all fields undergo encode/decode.
void Task::parseJSON(const std::string& line) {
  // Parse the whole thing.
  json::value* root = json::parse(line);
  if (root && root->type() == json::j_object) parseJSON((json::object*)root);

  delete root;
}

////////////////////////////////////////////////////////////////////////////////
void Task::parseJSON(const json::object* root_obj) {
  // For each object element...
  for (auto& i : root_obj->_data) {
    // If the attribute is a recognized column.
    std::string type = Task::attributes[i.first];
    if (type != "") {
      // Any specified id is ignored.
      if (i.first == "id")
        ;

      // Urgency, if present, is ignored.
      else if (i.first == "urgency")
        ;

      // TW-1274 Standardization.
      else if (i.first == "modification") {
        auto text = i.second->dump();
        Lexer::dequote(text);
        Datetime d(text);
        set("modified", d.toEpochString());
      }

      // Dates are converted from ISO to epoch.
      else if (type == "date") {
        auto text = i.second->dump();
        Lexer::dequote(text);
        Datetime d(text);
        set(i.first, text == "" ? "" : d.toEpochString());
      }

      // Tags are an array of JSON strings.
      else if (i.first == "tags" && i.second->type() == json::j_array) {
        auto tags = (json::array*)i.second;
        for (auto& t : tags->_data) {
          auto tag = (json::string*)t;
          addTag(tag->_data);
        }
      }

      // Dependencies can be exported as an array of strings.
      // 2016-02-21: This will be the only option in future releases.
      //             See other 2016-02-21 comments for details.
      else if (i.first == "depends" && i.second->type() == json::j_array) {
        auto deps = (json::array*)i.second;
        for (auto& t : deps->_data) {
          auto dep = (json::string*)t;
          addDependency(dep->_data);
        }
      }

      // Dependencies can be exported as a single comma-separated string.
      // 2016-02-21: Deprecated - see other 2016-02-21 comments for details.
      else if (i.first == "depends" && i.second->type() == json::j_string) {
        auto deps = (json::string*)i.second;

        // Fix for issue#2689: taskserver sometimes encodes the depends
        // property as a string of the format `[\"uuid\",\"uuid\"]`
        // The string includes the backslash-escaped `"` characters, making
        // it invalid JSON.  Since we know the characters we're looking for,
        // we'll just filter out everything else.
        std::string deps_str = deps->_data;
        if (deps_str.front() == '[' && deps_str.back() == ']') {
          std::string filtered;
          for (auto& c : deps_str) {
            if ((c >= '0' && c <= '9') || (c >= 'a' && c <= 'f') || c == ',' || c == '-') {
              filtered.push_back(c);
            }
          }
          deps_str = filtered;
        }
        auto uuids = split(deps_str, ',');

        for (const auto& uuid : uuids) addDependency(uuid);
      }

      // Strings are decoded.
      else if (type == "string") {
        auto text = i.second->dump();
        Lexer::dequote(text);
        set(i.first, json::decode(text));
      }

      // Other types are simply added.
      else {
        auto text = i.second->dump();
        Lexer::dequote(text);
        set(i.first, text);
      }
    }

    // UDA orphans and annotations do not have columns.
    else {
      // Annotations are an array of JSON objects with 'entry' and
      // 'description' values and must be converted.
      if (i.first == "annotations") {
        std::map<std::string, std::string> annos;

        // Fail if 'annotations' is not an array
        if (i.second->type() != json::j_array) {
          throw format("Annotations is malformed: {1}", i.second->dump());
        }

        auto atts = (json::array*)i.second;
        for (auto& annotations : atts->_data) {
          auto annotation = (json::object*)annotations;

          // Extract description. Fail if not present.
          auto what = (json::string*)annotation->_data["description"];
          if (!what) {
            annotation->_data.erase(
                "description");  // Erase NULL description inserted by failed lookup above
            throw format("Annotation is missing a description: {1}", annotation->dump());
          }

          // Extract 64-bit annotation entry value
          // Time travelers from 2038, we have your back.
          long long ann_timestamp;

          // Extract entry. Use current time if not present.
          auto when = (json::string*)annotation->_data["entry"];
          if (when)
            ann_timestamp = (long long)(Datetime(when->_data).toEpoch());
          else {
            annotation->_data.erase("entry");  // Erase NULL entry inserted by failed lookup above
            ann_timestamp = (long long)(Datetime().toEpoch());
          }

          std::stringstream name;
          name << "annotation_" << ann_timestamp;

          // Increment the entry timestamp in case of a conflict. Same
          // behaviour as CmdAnnotate.
          while (annos.find(name.str()) != annos.end()) {
            name.str("");  // Clear
            ann_timestamp++;
            name << "annotation_" << ann_timestamp;
          }

          annos.emplace(name.str(), json::decode(what->_data));
        }

        setAnnotations(annos);
      }

      // UDA Orphan - must be preserved.
      else {
#ifdef PRODUCT_TASKWARRIOR
        std::stringstream message;
        message << "Task::parseJSON found orphan '" << i.first << "' with value '" << i.second
                << "' --> preserved\n";
        Context::getContext().debug(message.str());
#endif
        auto text = i.second->dump();
        Lexer::dequote(text);
        set(i.first, json::decode(text));
      }
    }
  }
}

////////////////////////////////////////////////////////////////////////////////
// Note that all fields undergo encode/decode.
void Task::parseTC(rust::Box<tc::TaskData> task) {
  auto items = task->items();
  data.clear();
  for (auto& item : items) {
    data[static_cast<std::string>(item.prop)] = static_cast<std::string>(item.value);
  }

  // count annotations
  annotation_count = 0;
  for (auto i : data) {
    if (isAnnotationAttr(i.first)) {
      ++annotation_count;
    }
  }

  data["uuid"] = static_cast<std::string>(task->get_uuid().to_string());
  id = Context::getContext().tdb2.id(data["uuid"]);
}

////////////////////////////////////////////////////////////////////////////////
// No legacy formats are currently supported as of 2.4.0.
void Task::parseLegacy(const std::string& line) {
  switch (determineVersion(line)) {
    // File format version 1, from 2006-11-27 - 2007-12-31, v0.x+ - v0.9.3
    case 1:
      throw std::string(
          "Taskwarrior no longer supports file format 1, originally used between 27 November 2006 "
          "and 31 December 2007.");

    // File format version 2, from 2008-1-1 - 2009-3-23, v0.9.3 - v1.5.0
    case 2:
      throw std::string(
          "Taskwarrior no longer supports file format 2, originally used between 1 January 2008 "
          "and 12 April 2009.");

    // File format version 3, from 2009-3-23 - 2009-05-16, v1.6.0 - v1.7.1
    case 3:
      throw std::string(
          "Taskwarrior no longer supports file format 3, originally used between 23 March 2009 and "
          "16 May 2009.");

    // File format version 4, from 2009-05-16 - today, v1.7.1+
    case 4:
      break;

    default:
#ifdef PRODUCT_TASKWARRIOR
      std::stringstream message;
      message << "Invalid fileformat at line '" << line << '\'';
      Context::getContext().debug(message.str());
#endif
      throw std::string("Unrecognized Taskwarrior file format or blank line in data.");
      break;
  }

  recalc_urgency = true;
}

////////////////////////////////////////////////////////////////////////////////
std::string Task::composeJSON(bool decorate /*= false*/) const {
  std::stringstream out;
  out << '{';

  // ID inclusion is optional, but not a good idea, because it remains correct
  // only until the next gc.
  if (decorate) out << "\"id\":" << id << ',';

  // First the non-annotations.
  int attributes_written = 0;
  for (auto& i : data) {
    // Annotations are not written out here.
    if (!i.first.compare(0, 11, "annotation_", 11)) continue;

    // Tags and dependencies are handled below
    if (i.first == "tags" || isTagAttr(i.first)) continue;
    if (i.first == "depends" || isDepAttr(i.first)) continue;

    // If value is an empty string, do not ever output it
    if (i.second == "") continue;

    std::string type = Task::attributes[i.first];
    if (type == "") type = "string";

    // Date fields are written as ISO 8601.
    if (type == "date") {
      time_t epoch = get_date(i.first);
      if (epoch != 0) {
        Datetime d(i.second);
        if (attributes_written) out << ',';

        out << '"' << (i.first == "modification" ? "modified" : i.first)
            << "\":\""
            // Date was deleted, do not export parsed empty string
            << (i.second == "" ? "" : d.toISO()) << '"';

        ++attributes_written;
      }
    }

    /*
        else if (type == "duration")
        {
          // TODO Emit Datetime
        }
    */
    else if (type == "numeric") {
      if (attributes_written) out << ',';

      out << '"' << i.first << "\":" << i.second;

      ++attributes_written;
    }

    // Everything else is a quoted value.
    else {
      if (attributes_written) out << ',';

      out << '"' << i.first << "\":\"" << (type == "string" ? json::encode(i.second) : i.second)
          << '"';

      ++attributes_written;
    }
  }

  // Now the annotations, if any.
  if (annotation_count) {
    out << ',' << "\"annotations\":[";

    int annotations_written = 0;
    for (auto& i : data) {
      if (!i.first.compare(0, 11, "annotation_", 11)) {
        if (annotations_written) out << ',';

        Datetime d(i.first.substr(11));
        out << R"({"entry":")" << d.toISO() << R"(","description":")" << json::encode(i.second)
            << "\"}";

        ++annotations_written;
      }
    }

    out << ']';
  }

  auto tags = getTags();
  if (tags.size() > 0) {
    out << ',' << "\"tags\":[";

    int count = 0;
    for (const auto& tag : tags) {
      if (count++) out << ',';

      out << '"' << tag << '"';
    }

    out << ']';
    ++attributes_written;
  }

  auto depends = getDependencyUUIDs();
  if (depends.size() > 0) {
    out << ',' << "\"depends\":[";

    int count = 0;
    for (const auto& dep : depends) {
      if (count++) out << ',';

      out << '"' << dep << '"';
    }

    out << ']';
    ++attributes_written;
  }

#ifdef PRODUCT_TASKWARRIOR
  // Include urgency.
  if (decorate) out << ',' << "\"urgency\":" << urgency_c();
#endif

  out << '}';
  return out.str();
}

////////////////////////////////////////////////////////////////////////////////
int Task::getAnnotationCount() const {
  int count = 0;
  for (auto& ann : data)
    if (!ann.first.compare(0, 11, "annotation_", 11)) ++count;

  return count;
}

////////////////////////////////////////////////////////////////////////////////
bool Task::hasAnnotations() const { return annotation_count ? true : false; }

////////////////////////////////////////////////////////////////////////////////
// The timestamp is part of the name:
//    annotation_1234567890:"..."
//
// Note that the time is incremented (one second) in order to find a unique
// timestamp.
void Task::addAnnotation(const std::string& description) {
  time_t now = time(nullptr);
  std::string key;

  do {
    key = "annotation_" + format((long long int)now);
    ++now;
  } while (has(key));

  data[key] = json::decode(description);
  ++annotation_count;
  recalc_urgency = true;
}

////////////////////////////////////////////////////////////////////////////////
void Task::removeAnnotations() {
  // Erase old annotations.
  auto i = data.begin();
  while (i != data.end()) {
    if (!i->first.compare(0, 11, "annotation_", 11)) {
      --annotation_count;
      data.erase(i++);
    } else
      i++;
  }

  recalc_urgency = true;
}

////////////////////////////////////////////////////////////////////////////////
std::map<std::string, std::string> Task::getAnnotations() const {
  std::map<std::string, std::string> a;
  for (auto& ann : data)
    if (!ann.first.compare(0, 11, "annotation_", 11)) a.insert(ann);

  return a;
}

////////////////////////////////////////////////////////////////////////////////
void Task::setAnnotations(const std::map<std::string, std::string>& annotations) {
  // Erase old annotations.
  removeAnnotations();

  for (auto& anno : annotations) data.insert(anno);

  annotation_count = annotations.size();
  recalc_urgency = true;
}

#ifdef PRODUCT_TASKWARRIOR
////////////////////////////////////////////////////////////////////////////////
void Task::addDependency(int depid) {
  // Check that id is resolvable.
  std::string uuid = Context::getContext().tdb2.uuid(depid);
  if (uuid == "") throw format("Could not create a dependency on task {1} - not found.", depid);

  // the addDependency(&std::string) overload will check this, too, but here we
  // can give an more natural error message containing the id the user
  // provided.
  if (hasDependency(uuid)) {
    Context::getContext().footnote(format("Task {1} already depends on task {2}.", id, depid));
    return;
  }

  addDependency(uuid);
}
#endif

////////////////////////////////////////////////////////////////////////////////
void Task::addDependency(const std::string& uuid) {
  if (uuid == get("uuid")) throw std::string("A task cannot be dependent on itself.");

  if (hasDependency(uuid)) {
#ifdef PRODUCT_TASKWARRIOR
    Context::getContext().footnote(
        format("Task {1} already depends on task {2}.", get("uuid"), uuid));
#endif
    return;
  }

  // Store the dependency.
  set(dep2Attr(uuid), "x");

  // Prevent circular dependencies.
#ifdef PRODUCT_TASKWARRIOR
  if (dependencyIsCircular(*this))
    throw std::string("Circular dependency detected and disallowed.");
#endif

  recalc_urgency = true;
  fixDependsAttribute();
}

#ifdef PRODUCT_TASKWARRIOR
////////////////////////////////////////////////////////////////////////////////
void Task::removeDependency(int id) {
  std::string uuid = Context::getContext().tdb2.uuid(id);

  // The removeDependency(std::string&) method will check this too, but here we
  // can give a more natural error message containing the id provided by the user
  if (uuid == "" || !has(dep2Attr(uuid)))
    throw format("Could not delete a dependency on task {1} - not found.", id);
  removeDependency(uuid);
}

////////////////////////////////////////////////////////////////////////////////
void Task::removeDependency(const std::string& uuid) {
  auto depattr = dep2Attr(uuid);
  if (has(depattr))
    remove(depattr);
  else
    throw format("Could not delete a dependency on task {1} - not found.", uuid);

  recalc_urgency = true;
  fixDependsAttribute();
}

////////////////////////////////////////////////////////////////////////////////
bool Task::hasDependency(const std::string& uuid) const {
  auto depattr = dep2Attr(uuid);
  return has(depattr);
}

////////////////////////////////////////////////////////////////////////////////
std::vector<int> Task::getDependencyIDs() const {
  std::vector<int> ids;
  for (auto& attr : all()) {
    if (!isDepAttr(attr)) continue;
    auto dep = attr2Dep(attr);
    ids.push_back(Context::getContext().tdb2.id(dep));
  }

  return ids;
}

////////////////////////////////////////////////////////////////////////////////
std::vector<std::string> Task::getDependencyUUIDs() const {
  std::vector<std::string> uuids;
  for (auto& attr : all()) {
    if (!isDepAttr(attr)) continue;
    auto dep = attr2Dep(attr);
    uuids.push_back(dep);
  }

  return uuids;
}

////////////////////////////////////////////////////////////////////////////////
std::vector<Task> Task::getDependencyTasks() const {
  auto uuids = getDependencyUUIDs();

  // NOTE: this may seem inefficient, but note that `TDB2::get` performs a
  // linear search on each invocation, so scanning *once* is quite a bit more
  // efficient.
  std::vector<Task> blocking;
  if (uuids.size() > 0)
    for (auto& it : Context::getContext().tdb2.pending_tasks())
      if (it.getStatus() != Task::completed && it.getStatus() != Task::deleted &&
          std::find(uuids.begin(), uuids.end(), it.get("uuid")) != uuids.end())
        blocking.push_back(it);

  return blocking;
}

////////////////////////////////////////////////////////////////////////////////
std::vector<Task> Task::getBlockedTasks() const {
  auto uuid = get("uuid");

  std::vector<Task> blocked;
  for (auto& it : Context::getContext().tdb2.pending_tasks())
    if (it.getStatus() != Task::completed && it.getStatus() != Task::deleted &&
        it.hasDependency(uuid))
      blocked.push_back(it);

  return blocked;
}
#endif

////////////////////////////////////////////////////////////////////////////////
int Task::getTagCount() const {
  auto count = 0;
  for (auto& attr : data) {
    if (isTagAttr(attr.first)) {
      count++;
    }
  }
  return count;
}

////////////////////////////////////////////////////////////////////////////////
//
//              OVERDUE YESTERDAY DUE TODAY TOMORROW WEEK MONTH YEAR
// due:-1week      Y       -       -    -       -      ?    ?     ?
// due:-1day       Y       Y       -    -       -      ?    ?     ?
// due:today       Y       -       Y    Y       -      ?    ?     ?
// due:tomorrow    -       -       Y    -       Y      ?    ?     ?
// due:3days       -       -       Y    -       -      ?    ?     ?
// due:1month      -       -       -    -       -      -    -     ?
// due:1year       -       -       -    -       -      -    -     -
//
bool Task::hasTag(const std::string& tag) const {
  // Synthetic tags - dynamically generated, but do not occupy storage space.
  // Note: This list must match that in CmdInfo::execute.
  // Note: This list must match that in ::feedback_reserved_tags.
  if (isupper(tag[0])) {
    // NOTE: This list should be kept synchronized with:
    // * the list in CmdTags.cpp for the _tags command.
    // * the list in CmdInfo.cpp for the info command.
    if (tag == "BLOCKED") return is_blocked;
    if (tag == "UNBLOCKED") return !is_blocked;
    if (tag == "BLOCKING") return is_blocking;
#ifdef PRODUCT_TASKWARRIOR
    if (tag == "READY") return is_ready();
    if (tag == "DUE") return is_due();
    if (tag == "DUETODAY") return is_duetoday();  // 2016-03-29: Deprecated in 2.6.0
    if (tag == "TODAY") return is_duetoday();
    if (tag == "YESTERDAY") return is_dueyesterday();
    if (tag == "TOMORROW") return is_duetomorrow();
    if (tag == "OVERDUE") return is_overdue();
    if (tag == "WEEK") return is_dueweek();
    if (tag == "MONTH") return is_duemonth();
    if (tag == "QUARTER") return is_duequarter();
    if (tag == "YEAR") return is_dueyear();
#endif
    if (tag == "ACTIVE") return has("start");
    if (tag == "SCHEDULED") return has("scheduled");
    if (tag == "CHILD") return has("parent") || has("template");  // 2017-01-07: Deprecated in 2.6.0
    if (tag == "INSTANCE") return has("template") || has("parent");
    if (tag == "UNTIL") return has("until");
    if (tag == "ANNOTATED") return hasAnnotations();
    if (tag == "TAGGED") return getTagCount() > 0;
    if (tag == "PARENT") return has("mask") || has("last");  // 2017-01-07: Deprecated in 2.6.0
    if (tag == "TEMPLATE") return has("last") || has("mask");
    if (tag == "WAITING") return is_waiting();
    if (tag == "PENDING") return getStatus() == Task::pending;
    if (tag == "COMPLETED") return getStatus() == Task::completed;
    if (tag == "DELETED") return getStatus() == Task::deleted;
#ifdef PRODUCT_TASKWARRIOR
    if (tag == "UDA") return is_udaPresent();
    if (tag == "ORPHAN") return is_orphanPresent();
    if (tag == "LATEST") return id == Context::getContext().tdb2.latest_id();
#endif
    if (tag == "PROJECT") return has("project");
    if (tag == "PRIORITY") return has("priority");
  }

  // Concrete tags.
  if (has(tag2Attr(tag))) return true;

  return false;
}

////////////////////////////////////////////////////////////////////////////////
void Task::addTag(const std::string& tag) {
  auto attr = tag2Attr(tag);
  if (!has(attr)) {
    set(attr, "x");
    recalc_urgency = true;
    fixTagsAttribute();
  }
}

////////////////////////////////////////////////////////////////////////////////
void Task::setTags(const std::vector<std::string>& tags) {
  auto existing = getTags();

  // edit in-place, determining which should be
  // added and which should be removed
  std::vector<std::string> toAdd;
  std::vector<std::string> toRemove;

  for (auto& tag : tags) {
    if (std::find(existing.begin(), existing.end(), tag) == existing.end()) toAdd.push_back(tag);
  }

  for (auto& tag : getTags()) {
    if (std::find(tags.begin(), tags.end(), tag) == tags.end()) {
      toRemove.push_back(tag);
    }
  }

  for (auto& tag : toRemove) {
    removeTag(tag);
  }
  for (auto& tag : toAdd) {
    addTag(tag);
  }

  // (note: addTag / removeTag took care of recalculating urgency)
}

////////////////////////////////////////////////////////////////////////////////
std::vector<std::string> Task::getTags() const {
  std::vector<std::string> tags;

  for (auto& attr : data) {
    if (!isTagAttr(attr.first)) {
      continue;
    }
    auto tag = attr2Tag(attr.first);
    tags.push_back(tag);
  }

  return tags;
}

////////////////////////////////////////////////////////////////////////////////
void Task::removeTag(const std::string& tag) {
  auto attr = tag2Attr(tag);
  if (has(attr)) {
    data.erase(attr);
    recalc_urgency = true;
    fixTagsAttribute();
  }
}

////////////////////////////////////////////////////////////////////////////////
void Task::fixTagsAttribute() {
  // Fix up the old `tags` attribute to match the `tag_..` attributes (or
  // remove it if there are no tags)
  auto tags = getTags();
  if (tags.size() > 0) {
    set("tags", join(",", tags));
  } else {
    remove("tags");
  }
}

////////////////////////////////////////////////////////////////////////////////
bool Task::isTagAttr(const std::string& attr) { return attr.compare(0, 4, "tag_") == 0; }

////////////////////////////////////////////////////////////////////////////////
std::string Task::tag2Attr(const std::string& tag) {
  std::stringstream tag_attr;
  tag_attr << "tag_" << tag;
  return tag_attr.str();
}

////////////////////////////////////////////////////////////////////////////////
std::string Task::attr2Tag(const std::string& attr) {
  assert(isTagAttr(attr));
  return attr.substr(4);
}

////////////////////////////////////////////////////////////////////////////////
void Task::fixDependsAttribute() {
  // Fix up the old `depends` attribute to match the `dep_..` attributes (or
  // remove it if there are no deps)
  auto deps = getDependencyUUIDs();
  if (deps.size() > 0) {
    set("depends", join(",", deps));
  } else {
    remove("depends");
  }
}

////////////////////////////////////////////////////////////////////////////////
bool Task::isDepAttr(const std::string& attr) { return attr.compare(0, 4, "dep_") == 0; }

////////////////////////////////////////////////////////////////////////////////
std::string Task::dep2Attr(const std::string& tag) {
  std::stringstream tag_attr;
  tag_attr << "dep_" << tag;
  return tag_attr.str();
}

////////////////////////////////////////////////////////////////////////////////
std::string Task::attr2Dep(const std::string& attr) {
  assert(isDepAttr(attr));
  return attr.substr(4);
}

////////////////////////////////////////////////////////////////////////////////
bool Task::isAnnotationAttr(const std::string& attr) {
  return attr.compare(0, 11, "annotation_") == 0;
}

#ifdef PRODUCT_TASKWARRIOR
////////////////////////////////////////////////////////////////////////////////
// A UDA Orphan is an attribute that is not represented in context.columns.
std::vector<std::string> Task::getUDAOrphans() const {
  std::vector<std::string> orphans;
  for (auto& it : data)
    if (Context::getContext().columns.find(it.first) == Context::getContext().columns.end())
      if (not(isAnnotationAttr(it.first) || isTagAttr(it.first) || isDepAttr(it.first)))
        orphans.push_back(it.first);

  return orphans;
}

////////////////////////////////////////////////////////////////////////////////
void Task::substitute(const std::string& from, const std::string& to, const std::string& flags) {
  bool global = (flags.find('g') != std::string::npos ? true : false);

  // Get the data to modify.
  std::string description = get("description");
  auto annotations = getAnnotations();

  // Count the changes, so we know whether to proceed to annotations, after
  // modifying description.
  int changes = 0;
  bool done = false;

  // Regex support is optional.
  if (Task::regex) {
    // Create the regex.
    RX rx(from, Task::searchCaseSensitive);
    std::vector<int> start;
    std::vector<int> end;

    // Perform all subs on description.
    if (rx.match(start, end, description)) {
      int skew = 0;
      for (unsigned int i = 0; i < start.size() && !done; ++i) {
        description.replace(start[i] + skew, end[i] - start[i], to);
        skew += to.length() - (end[i] - start[i]);
        ++changes;

        if (!global) done = true;
      }
    }

    if (!done) {
      // Perform all subs on annotations.
      for (auto& it : annotations) {
        start.clear();
        end.clear();
        if (rx.match(start, end, it.second)) {
          int skew = 0;
          for (unsigned int i = 0; i < start.size() && !done; ++i) {
            it.second.replace(start[i + skew], end[i] - start[i], to);
            skew += to.length() - (end[i] - start[i]);
            ++changes;

            if (!global) done = true;
          }
        }
      }
    }
  } else {
    // Perform all subs on description.
    int counter = 0;
    std::string::size_type pos = 0;
    int skew = 0;

    while ((pos = ::find(description, from, pos, Task::searchCaseSensitive)) != std::string::npos &&
           !done) {
      description.replace(pos + skew, from.length(), to);
      skew += to.length() - from.length();

      pos += to.length();
      ++changes;

      if (!global) done = true;

      if (++counter > APPROACHING_INFINITY)
        throw format(
            "Terminated substitution because more than {1} changes were made - infinite loop "
            "protection.",
            APPROACHING_INFINITY);
    }

    if (!done) {
      // Perform all subs on annotations.
      counter = 0;
      for (auto& anno : annotations) {
        pos = 0;
        skew = 0;
        while ((pos = ::find(anno.second, from, pos, Task::searchCaseSensitive)) !=
                   std::string::npos &&
               !done) {
          anno.second.replace(pos + skew, from.length(), to);
          skew += to.length() - from.length();

          pos += to.length();
          ++changes;

          if (!global) done = true;

          if (++counter > APPROACHING_INFINITY)
            throw format(
                "Terminated substitution because more than {1} changes were made - infinite loop "
                "protection.",
                APPROACHING_INFINITY);
        }
      }
    }
  }

  if (changes) {
    set("description", description);
    setAnnotations(annotations);
    recalc_urgency = true;
  }
}
#endif

////////////////////////////////////////////////////////////////////////////////
// Validate a task for addition, raising user-visible errors for inconsistent or
// incorrect inputs. This is called before `Task::validate`.
void Task::validate_add() {
  // There is no fixing a missing description.
  if (!has("description"))
    throw std::string("A task must have a description.");
  else if (get("description") == "")
    throw std::string("Cannot add a task that is blank.");

  // Cannot have an old-style recur frequency with no due date - when would it recur?
  if (has("recur") && (!has("due") || get("due") == ""))
    throw std::string("A recurring task must also have a 'due' date.");
}

////////////////////////////////////////////////////////////////////////////////
// The purpose of Task::validate is three-fold:
//   1) To provide missing attributes where possible
//   2) To provide suitable warnings about odd states
//   3) To update status depending on other attributes
//
// As required by TaskChampion, no combination of properties and values is an
// error. This function will try to make sensible defaults and resolve inconsistencies.
// Critically, note that despite the name this is not a read-only function.
//
void Task::validate(bool applyDefault /* = true */) {
  Task::status status = Task::pending;
  if (get("status") != "") status = getStatus();

  // 1) Provide missing attributes where possible
  // Provide a UUID if necessary. Validate if present.
  std::string uid = get("uuid");
  if (has("uuid") && uid != "") {
    Lexer lex(uid);
    std::string token;
    Lexer::Type type;
    // `uuid` is not a property in the TaskChampion model, so an invalid UUID is
    // actually an error.
    if (!lex.isUUID(token, type, true)) throw format("Not a valid UUID '{1}'.", uid);
  } else
    set("uuid", uuid());

  // TODO Obsolete remove for 3.0.0
  // Recurring tasks get a special status.
  if (status == Task::pending && has("due") && has("recur") &&
      (!has("parent") || get("parent") == "") && (!has("template") || get("template") == "")) {
    status = Task::recurring;
  }
  /*
    // TODO Add for 3.0.0
    if (status == Task::pending &&
        has ("due")             &&
        has ("recur")          &&
        (! has ("template") || get ("template") == ""))
    {
      status = Task::recurring;
    }
  */

  // Tasks with a wait: date get a special status.
  else if (status == Task::pending && has("wait") && get("wait") != "")
    status = Task::waiting;

  // By default, tasks are pending.
  else if (!has("status") || get("status") == "")
    status = Task::pending;

  // Default to 'periodic' type recurrence.
  if (status == Task::recurring && (!has("rtype") || get("rtype") == "")) {
    set("rtype", "periodic");
  }

  // Store the derived status.
  setStatus(status);

#ifdef PRODUCT_TASKWARRIOR
  // Provide an entry date unless user already specified one.
  if (!has("entry") || get("entry") == "") setAsNow("entry");

  // Completed tasks need an end date, so inherit the entry date.
  if ((status == Task::completed || status == Task::deleted) && (!has("end") || get("end") == ""))
    setAsNow("end");

  // Pending tasks cannot have an end date, remove if present
  if ((status == Task::pending) && (get("end") != "")) remove("end");

  // Provide a modified date unless user already specified one.
  if (!has("modified") || get("modified") == "") setAsNow("modified");

  if (applyDefault && (!has("parent") || get("parent") == "")) {
    // Override with default.project, if not specified.
    if (Task::defaultProject != "" && !has("project")) {
      if (Context::getContext().columns["project"]->validate(Task::defaultProject))
        set("project", Task::defaultProject);
    }

    // Override with default.due, if not specified.
    if (Task::defaultDue != "" && !has("due")) {
      if (Context::getContext().columns["due"]->validate(Task::defaultDue)) {
        Duration dur(Task::defaultDue);
        if (dur.toTime_t() != 0)
          set("due", (Datetime() + dur.toTime_t()).toEpoch());
        else
          set("due", Datetime(Task::defaultDue).toEpoch());
      }
    }

    // Override with default.scheduled, if not specified.
    if (Task::defaultScheduled != "" && !has("scheduled")) {
      if (Context::getContext().columns["scheduled"]->validate(Task::defaultScheduled)) {
        Duration dur(Task::defaultScheduled);
        if (dur.toTime_t() != 0)
          set("scheduled", (Datetime() + dur.toTime_t()).toEpoch());
        else
          set("scheduled", Datetime(Task::defaultScheduled).toEpoch());
      }
    }

    // If a UDA has a default value in the configuration,
    // override with uda.(uda).default, if not specified.
    // Gather a list of all UDAs with a .default value
    std::vector<std::string> udas;
    for (auto& var : Context::getContext().config) {
      if (!var.first.compare(0, 4, "uda.", 4) && var.first.find(".default") != std::string::npos) {
        auto period = var.first.find('.', 4);
        if (period != std::string::npos) udas.push_back(var.first.substr(4, period - 4));
      }
    }

    if (udas.size()) {
      // For each of those, setup the default value on the task now,
      // of course only if we don't have one on the command line already
      for (auto& uda : udas) {
        std::string defVal = Context::getContext().config.get("uda." + uda + ".default");

        // If the default is empty, or we already have a value, skip it
        if (defVal != "" && get(uda) == "") set(uda, defVal);
      }
    }
  }
#endif

  // 2) To provide suitable warnings about odd states

  // Date relationships.
  validate_before("wait", "due");
  validate_before("entry", "start");
  validate_before("entry", "end");
  validate_before("wait", "scheduled");
  validate_before("scheduled", "start");
  validate_before("scheduled", "due");
  validate_before("scheduled", "end");

  if (!has("description") || get("description") == "")
    Context::getContext().footnote(format("Warning: task has no description."));

  // Cannot have an old-style recur frequency with no due date - when would it recur?
  if (has("recur") && (!has("due") || get("due") == "")) {
    Context::getContext().footnote(format("Warning: recurring task has no due date."));
    remove("recur");
  }

  // Old-style recur durations must be valid.
  if (has("recur")) {
    std::string value = get("recur");
    if (value != "") {
      Duration p;
      std::string::size_type i = 0;
      if (!p.parse(value, i)) {
        // TODO Ideal location to map unsupported old recurrence periods to supported values.
        Context::getContext().footnote(
            format("Warning: The recurrence value '{1}' is not valid.", value));
        remove("recur");
      }
    }
  }
}

////////////////////////////////////////////////////////////////////////////////
void Task::validate_before(const std::string& left, const std::string& right) {
#ifdef PRODUCT_TASKWARRIOR
  if (has(left) && has(right)) {
    Datetime date_left(get_date(left));
    Datetime date_right(get_date(right));

    // if date is zero, then it is being removed (e.g. "due: wait:1day")
    if (date_left > date_right && date_right.toEpoch() != 0)
      Context::getContext().footnote(format(
          "Warning: You have specified that the '{1}' date is after the '{2}' date.", left, right));
  }
#endif
}

////////////////////////////////////////////////////////////////////////////////
// Encode values prior to serialization.
//   [  -> &open;
//   ]  -> &close;
const std::string Task::encode(const std::string& value) const {
  auto modified = str_replace(value, "[", "&open;");
  return str_replace(modified, "]", "&close;");
}

////////////////////////////////////////////////////////////////////////////////
// Decode values after parse.
//   [  <- &open;
//   ]  <- &close;
const std::string Task::decode(const std::string& value) const {
  if (value.find('&') == std::string::npos) return value;

  auto modified = str_replace(value, "&open;", "[");
  return str_replace(modified, "&close;", "]");
}

////////////////////////////////////////////////////////////////////////////////
int Task::determineVersion(const std::string& line) {
  // Version 2 looks like:
  //
  //   uuid status [tags] [attributes] description\n
  //
  // Where uuid looks like:
  //
  //   27755d92-c5e9-4c21-bd8e-c3dd9e6d3cf7
  //
  // Scan for the hyphens in the uuid, the following space, and a valid status
  // character.
  if (line[8] == '-' && line[13] == '-' && line[18] == '-' && line[23] == '-' && line[36] == ' ' &&
      (line[37] == '-' || line[37] == '+' || line[37] == 'X' || line[37] == 'r')) {
    // Version 3 looks like:
    //
    //   uuid status [tags] [attributes] [annotations] description\n
    //
    // Scan for the number of [] pairs.
    auto tagAtts = line.find("] [", 0);
    auto attsAnno = line.find("] [", tagAtts + 1);
    auto annoDesc = line.find("] ", attsAnno + 1);
    if (tagAtts != std::string::npos && attsAnno != std::string::npos &&
        annoDesc != std::string::npos)
      return 3;
    else
      return 2;
  }

  // Version 4 looks like:
  //
  //   [name:"value" ...]
  //
  // Scan for [, ] and :".
  else if (line[0] == '[' && line[line.length() - 1] == ']' &&
           line.find("uuid:\"") != std::string::npos)
    return 4;

  // Version 1 looks like:
  //
  //   [tags] [attributes] description\n
  //   X [tags] [attributes] description\n
  //
  // Scan for the first character being either the bracket or X.
  else if (line.find("X [") == 0 ||
           (line[0] == '[' && line.substr(line.length() - 1, 1) != "]" && line.length() > 3))
    return 1;

  // Version 5?
  //
  // Fortunately, with the hindsight that will come with version 5, the
  // identifying characteristics of 1, 2, 3 and 4 may be modified such that if 5
  // has a UUID followed by a status, then there is still a way to differentiate
  // between 2, 3, 4 and 5.
  //
  // The danger is that a version 3 binary reads and misinterprets a version 4
  // file.  This is why it is a good idea to rely on an explicit version
  // declaration rather than chance positioning.

  // Zero means 'no idea'.
  return 0;
}

////////////////////////////////////////////////////////////////////////////////
// Urgency is defined as a polynomial, the value of which is calculated in this
// function, according to:
//
//   U = A.t  + B.t  + C.t  ...
//          a      b      c
//
//   U       = urgency
//   A       = coefficient for term a
//   t sub a = numeric scale from 0 -> 1, with 1 being the highest
//             urgency, derived from one task attribute and mapped
//             to the numeric scale
//
// See rfc31-urgency.txt for full details.
//
float Task::urgency_c() const {
  float value = 0.0;
#ifdef PRODUCT_TASKWARRIOR
  value += fabsf(Task::urgencyProjectCoefficient) > epsilon
               ? (urgency_project() * Task::urgencyProjectCoefficient)
               : 0.0;
  value += fabsf(Task::urgencyActiveCoefficient) > epsilon
               ? (urgency_active() * Task::urgencyActiveCoefficient)
               : 0.0;
  value += fabsf(Task::urgencyScheduledCoefficient) > epsilon
               ? (urgency_scheduled() * Task::urgencyScheduledCoefficient)
               : 0.0;
  value += fabsf(Task::urgencyWaitingCoefficient) > epsilon
               ? (urgency_waiting() * Task::urgencyWaitingCoefficient)
               : 0.0;
  value += fabsf(Task::urgencyBlockedCoefficient) > epsilon
               ? (urgency_blocked() * Task::urgencyBlockedCoefficient)
               : 0.0;
  value += fabsf(Task::urgencyAnnotationsCoefficient) > epsilon
               ? (urgency_annotations() * Task::urgencyAnnotationsCoefficient)
               : 0.0;
  value += fabsf(Task::urgencyTagsCoefficient) > epsilon
               ? (urgency_tags() * Task::urgencyTagsCoefficient)
               : 0.0;
  value += fabsf(Task::urgencyDueCoefficient) > epsilon
               ? (urgency_due() * Task::urgencyDueCoefficient)
               : 0.0;
  value += fabsf(Task::urgencyBlockingCoefficient) > epsilon
               ? (urgency_blocking() * Task::urgencyBlockingCoefficient)
               : 0.0;
  value += fabsf(Task::urgencyAgeCoefficient) > epsilon
               ? (urgency_age() * Task::urgencyAgeCoefficient)
               : 0.0;

  const std::string taskProjectName = get("project");
  // Tag- and project-specific coefficients.
  for (auto& var : Task::coefficients) {
    if (fabs(var.second) > epsilon) {
      if (!var.first.compare(0, 13, "urgency.user.", 13)) {
        // urgency.user.project.<project>.coefficient
        auto end = std::string::npos;
        if (var.first.substr(13, 8) == "project." &&
            (end = var.first.find(".coefficient")) != std::string::npos) {
          std::string project = var.first.substr(21, end - 21);

          if (taskProjectName == project || taskProjectName.find(project + '.') == 0) {
            value += var.second;
          }
        }

        // urgency.user.tag.<tag>.coefficient
        if (var.first.substr(13, 4) == "tag." &&
            (end = var.first.find(".coefficient")) != std::string::npos) {
          std::string tag = var.first.substr(17, end - 17);

          if (hasTag(tag)) value += var.second;
        }

        // urgency.user.keyword.<keyword>.coefficient
        if (var.first.substr(13, 8) == "keyword." &&
            (end = var.first.find(".coefficient")) != std::string::npos) {
          std::string keyword = var.first.substr(21, end - 21);

          if (get("description").find(keyword) != std::string::npos) value += var.second;
        }
      } else if (var.first.substr(0, 12) == "urgency.uda.") {
        // urgency.uda.<name>.coefficient
        // urgency.uda.<name>.<value>.coefficient
        auto end = var.first.find(".coefficient");
        if (end != std::string::npos) {
          const std::string uda = var.first.substr(12, end - 12);
          auto dot = uda.find('.');
          if (dot == std::string::npos) {
            // urgency.uda.<name>.coefficient
            if (has(uda)) value += var.second;
          } else {
            // urgency.uda.<name>.<value>.coefficient
            if (get(uda.substr(0, dot)) == uda.substr(dot + 1)) value += var.second;
          }
        }
      }
    }
  }

  if (is_blocking && Context::getContext().config.getBoolean("urgency.inherit")) {
    float prev = value;
    value = std::max(value, urgency_inherit());

    // This is a hackish way of making sure parent tasks are sorted above
    // child tasks.  For reports that hide blocked tasks, this is not needed.
    if (prev <= value) value += 0.01;
  }
#endif

  return value;
}

////////////////////////////////////////////////////////////////////////////////
float Task::urgency() {
  if (recalc_urgency) {
    urgency_value = urgency_c();

    // Return the sum of all terms.
    recalc_urgency = false;
  }

  return urgency_value;
}

////////////////////////////////////////////////////////////////////////////////
float Task::urgency_inherit() const {
  float v = -FLT_MAX;
#ifdef PRODUCT_TASKWARRIOR
  // Calling getBlockedTasks is rather expensive.
  // It is called recursively for each dependency in the chain here.
  for (auto& task : getBlockedTasks()) {
    // Find highest urgency in all blocked tasks.
    v = std::max(v, task.urgency());
  }
#endif

  return v;
}

////////////////////////////////////////////////////////////////////////////////
float Task::urgency_project() const {
  if (has("project")) return 1.0;

  return 0.0;
}

////////////////////////////////////////////////////////////////////////////////
float Task::urgency_active() const {
  if (has("start")) return 1.0;

  return 0.0;
}

////////////////////////////////////////////////////////////////////////////////
float Task::urgency_scheduled() const {
  if (has("scheduled") && get_date("scheduled") < time(nullptr)) return 1.0;

  return 0.0;
}

////////////////////////////////////////////////////////////////////////////////
float Task::urgency_waiting() const {
  if (is_waiting()) return 1.0;

  return 0.0;
}

////////////////////////////////////////////////////////////////////////////////
// A task is blocked only if the task it depends upon is pending/waiting.
float Task::urgency_blocked() const {
  if (is_blocked) return 1.0;

  return 0.0;
}

////////////////////////////////////////////////////////////////////////////////
float Task::urgency_annotations() const {
  if (annotation_count >= 3)
    return 1.0;
  else if (annotation_count == 2)
    return 0.9;
  else if (annotation_count == 1)
    return 0.8;

  return 0.0;
}

////////////////////////////////////////////////////////////////////////////////
float Task::urgency_tags() const {
  switch (getTagCount()) {
    case 0:
      return 0.0;
    case 1:
      return 0.8;
    case 2:
      return 0.9;
    default:
      return 1.0;
  }
}

////////////////////////////////////////////////////////////////////////////////
//
//     Past                  Present                              Future
//     Overdue               Due                                     Due
//
//     -7 -6 -5 -4 -3 -2 -1  0  1  2  3  4  5  6  7  8  9 10 11 12 13 14 days
//
// <-- 1.0                         linear                            0.2 -->
//     capped                                                        capped
//
//
float Task::urgency_due() const {
  if (has("due")) {
    Datetime now;
    Datetime due(get_date("due"));

    // Map a range of 21 days to the value 0.2 - 1.0
    float days_overdue = (now - due) / 86400.0;
    if (days_overdue >= 7.0)
      return 1.0;  // < 1 wk ago
    else if (days_overdue >= -14.0)
      return ((days_overdue + 14.0) * 0.8 / 21.0) + 0.2;
    else
      return 0.2;  // > 2 wks
  }

  return 0.0;
}

////////////////////////////////////////////////////////////////////////////////
float Task::urgency_age() const {
  if (!has("entry")) return 1.0;

  Datetime now;
  Datetime entry(get_date("entry"));
  int age = (now - entry) / 86400;  // in days

  if (Task::urgencyAgeMax == 0 || age > Task::urgencyAgeMax) return 1.0;

  return (1.0 * age / Task::urgencyAgeMax);
}

////////////////////////////////////////////////////////////////////////////////
float Task::urgency_blocking() const {
  if (is_blocking) return 1.0;

  return 0.0;
}

#ifdef PRODUCT_TASKWARRIOR
////////////////////////////////////////////////////////////////////////////////
// Arguably does not belong here. This method reads the parse tree and calls
// Task methods. It could be a standalone function with no loss in access, as
// well as reducing the object depdendencies of Task.
//
// It came from the Command base object, but doesn't really belong there either.
void Task::modify(modType type, bool text_required /* = false */) {
  std::string label = "  MODIFICATION ";

  // while reading the parse tree, consider DOM references in the context of
  // this task
  auto currentTask = Context::getContext().withCurrentTask(this);

  // Need this for later comparison.
  auto originalStatus = getStatus();

  std::string text = "";
  bool mods = false;
  for (auto& a : Context::getContext().cli2._args) {
    if (a.hasTag("MODIFICATION")) {
      if (a._lextype == Lexer::Type::pair) {
        // 'canonical' is the canonical name. Needs to be said.
        // 'value' requires eval.
        std::string name = a.attribute("canonical");
        std::string value = a.attribute("value");
        if (value == "" || value == "''" || value == "\"\"") {
          // Special case: Handle bulk removal of 'tags' and 'depends" virtual
          // attributes
          if (name == "depends") {
            for (auto dep : getDependencyUUIDs()) removeDependency(dep);
          } else if (name == "tags") {
            for (auto tag : getTags()) removeTag(tag);
          }

          // ::composeF4 will skip if the value is blank, but the presence of
          // the attribute will prevent ::validate from applying defaults.
          if ((has(name) && get(name) != "") ||
              (name == "due" && Context::getContext().config.has("default.due")) ||
              (name == "scheduled" && Context::getContext().config.has("default.scheduled")) ||
              (name == "project" && Context::getContext().config.has("default.project"))) {
            mods = true;
            set(name, "");
          }

          Context::getContext().debug(label + name + " <-- ''");
        } else {
          Lexer::dequote(value);

          // Get the column info. Some columns are not modifiable.
          Column* column = Context::getContext().columns[name];
          if (!column || !column->modifiable())
            throw format("The '{1}' attribute does not allow a value of '{2}'.", name, value);

          // Delegate modification to the column object or their base classes.
          if (name == "depends" || name == "tags" || name == "recur" || column->type() == "date" ||
              column->type() == "duration" || column->type() == "numeric" ||
              column->type() == "string" || column->type() == "uuid") {
            column->modify(*this, value);
            mods = true;
          }

          else
            throw format("Unrecognized column type '{1}' for column '{2}'", column->type(), name);
        }
      }

      // Perform description/annotation substitution.
      else if (a._lextype == Lexer::Type::substitution) {
        Context::getContext().debug(label + "substitute " + a.attribute("raw"));
        substitute(a.attribute("from"), a.attribute("to"), a.attribute("flags"));
        mods = true;
      }

      // Tags need special handling because they are essentially a vector stored
      // in a single string, therefore Task::{add,remove}Tag must be called as
      // appropriate.
      else if (a._lextype == Lexer::Type::tag) {
        std::string tag = a.attribute("name");
        feedback_reserved_tags(tag);

        if (a.attribute("sign") == "+") {
          Context::getContext().debug(label + "tags <-- add '" + tag + '\'');
          addTag(tag);
          feedback_special_tags(*this, tag);
        } else {
          Context::getContext().debug(label + "tags <-- remove '" + tag + '\'');
          removeTag(tag);
        }

        mods = true;
      }

      // Unknown args are accumulated as though they were WORDs.
      else {
        if (text != "") text += ' ';
        text += a.attribute("raw");
      }
    }
  }

  // Task::modType determines what happens to the WORD arguments, if there are
  //  any.
  if (text != "") {
    Lexer::dequote(text);

    switch (type) {
      case modReplace:
        Context::getContext().debug(label + "description <-- '" + text + '\'');
        set("description", text);
        break;

      case modPrepend:
        Context::getContext().debug(label + "description <-- '" + text + "' + description");
        set("description", text + ' ' + get("description"));
        break;

      case modAppend:
        Context::getContext().debug(label + "description <-- description + '" + text + '\'');
        set("description", get("description") + ' ' + text);
        break;

      case modAnnotate:
        Context::getContext().debug(label + "new annotation <-- '" + text + '\'');
        addAnnotation(text);
        break;
    }
  } else if (!mods && text_required)
    throw std::string("Additional text must be provided.");

  // Modifying completed/deleted tasks generates a message, if the modification
  // does not change status.
  if ((getStatus() == Task::completed || getStatus() == Task::deleted) &&
      getStatus() == originalStatus) {
    auto uuid = get("uuid").substr(0, 8);
    Context::getContext().footnote(
        format("Note: Modified task {1} is {2}. You may wish to make this task pending with: task "
               "{3} modify status:pending",
               uuid, get("status"), uuid));
  }
}
#endif

////////////////////////////////////////////////////////////////////////////////
// Compare this task to another and summarize the differences for display, in
// the future tense ("Foo will be set to ..").
std::string Task::diff(const Task& after) const {
  // Attributes are all there is, so figure the different attribute names
  // between this (before) and after.
  std::vector<std::string> beforeAtts;
  for (auto& att : data) beforeAtts.push_back(att.first);

  std::vector<std::string> afterAtts;
  for (auto& att : after.data) afterAtts.push_back(att.first);

  std::vector<std::string> beforeOnly;
  std::vector<std::string> afterOnly;
  listDiff(beforeAtts, afterAtts, beforeOnly, afterOnly);

  // Now start generating a description of the differences.
  std::stringstream out;
  for (auto& name : beforeOnly) {
    if (isAnnotationAttr(name)) {
      out << "  - " << format("Annotation {1} will be removed.", name) << "\n";
    } else if (isTagAttr(name)) {
      out << "  - " << format("Tag {1} will be removed.", attr2Tag(name)) << "\n";
    } else if (isDepAttr(name)) {
      out << "  - " << format("Depenency on {1} will be removed.", attr2Dep(name)) << "\n";
    } else if (name == "depends" || name == "tags") {
      // do nothing for legacy attributes
    } else {
      out << "  - " << format("{1} will be deleted.", Lexer::ucFirst(name)) << "\n";
    }
  }

  for (auto& name : afterOnly) {
    if (isAnnotationAttr(name)) {
      out << format("Annotation of {1} will be added.\n", after.get(name));
    } else if (isTagAttr(name)) {
      out << format("Tag {1} will be added.\n", attr2Tag(name));
    } else if (isDepAttr(name)) {
      out << format("Dependency on {1} will be added.\n", attr2Dep(name));
    } else if (name == "depends" || name == "tags") {
      // do nothing for legacy attributes
    } else
      out << "  - "
          << format("{1} will be set to '{2}'.", Lexer::ucFirst(name),
                    renderAttribute(name, after.get(name)))
          << "\n";
  }

  for (auto& name : beforeAtts) {
    // Ignore UUID differences, and find values that changed, but are not also
    // in the beforeOnly and afterOnly lists, which have been handled above..
    if (name != "uuid" && get(name) != after.get(name) &&
        std::find(beforeOnly.begin(), beforeOnly.end(), name) == beforeOnly.end() &&
        std::find(afterOnly.begin(), afterOnly.end(), name) == afterOnly.end()) {
      if (name == "depends" || name == "tags") {
        // do nothing for legacy attributes
      } else if (isTagAttr(name) || isDepAttr(name)) {
        // ignore new attributes
      } else if (isAnnotationAttr(name)) {
        out << format("Annotation will be changed to {1}.\n", after.get(name));
      } else
        out << "  - "
            << format("{1} will be changed from '{2}' to '{3}'.", Lexer::ucFirst(name),
                      renderAttribute(name, get(name)), renderAttribute(name, after.get(name)))
            << "\n";
    }
  }

  // Shouldn't just say nothing.
  if (out.str().length() == 0) out << "  - No changes will be made.\n";

  return out.str();
}

////////////////////////////////////////////////////////////////////////////////