JIRA Cycle Data extract utility

NOTE: Whilst this tool should remain functional, development has moved onto its leaner, meaner, and more powerful offspring, jira-agile-metrics. See https://github.com/DeloitteDigitalUK/jira-agile-metrics

This utility helps extract data from JIRA for processing with the ActionableAgile™ Analytics tool (https://www.actionableagile.com/analytics-tools/), as well as ad-hoc analysis using Excel.

It will produce a CSV file with one row for each JIRA issue matching a set of filter criteria, containing basic information about the issue as well as the date the issue entered each step in the main cycle workflow.

This data can be used to produce a Cumulative Flow Diagram, a cycle time scatterplot, a cycle time histogram, and other analytics based on cycle time.

To make it easier to draw these diagrams, the tool can also be used to output CSV files with pre-calculated values for charting in e.g. Excel.

Finally, if you have the right dependencies installed, it can output basic charts as images.

Installation

Install Python 2.7 and pip. See http://pip.readthedocs.org/en/stable/installing/.

Install using pip:

$ pip install jira-cycle-extract

If you get errors, try to install numpy and pandas separately first:

$ pip install numpy pandas
$ pip install jira-cycle-extract

This will install a binary called jira-cycle-extract. You can test that it was correctly installed using:

$ jira-cycle-extract --help

If this doesn't work, check the output of pip install jira-cycle-extract to see where it may have installed the binary.

To use the built-in charting capabilities, you need to install Seaborn (which in turn installs Matplotlib and SciPy) and Statsmodels. You can get these with the charting extra:

$ pip install jira-cycle-extract[charting]

These dependencies are not installed by default because they can sometimes be a bit tricky to install.

Using Docker

If you have Docker installed, you can run jira-cycle-extract as a docker image:

$ docker run --rm -it -v "$PWD":/data optilude/jira-cycle-extract:latest --help

This will map the working directory (/data) of the containerised application to the current working directory ($PWD). Any files you specify on the command line (without any further path prefixes) will be read from or output to the current directory, e.g:

$ docker run --rm -it -v "$PWD":/data optilude/jira-cycle-extract:latest config.yml cycle.csv

Configuration

Write a YAML configuration file like so, calling it e.g. config.yaml:

# How to connect to JIRA?
Connection:
    Domain: https://myserver.atlassian.net
    Username: myusername # If missing, you will be prompted at runtime
    Password: secret     # If missing, you will be prompted at runtime
    Jira:                #List of key / value pairs that will be passed as options to jira as per https://jira.readthedocs.io/en/master/api.html#jira

# What to search for?
Criteria:
    Project: ABC # JIRA project key to search
    Issue types: # Which issue types to include
        - Story
        - Defect
    Valid resolutions: # Which resolution statuses to include (unresolved is always included)
        - Done
        - Closed
    JQL: labels != "Spike" # Additional filter as raw JQL, optional

# Describe the workflow. Each step can be mapped to either a single JIRA
# status, or a list of statuses that will be treated as equivalent
Workflow:
    Open: Open
    Analysis IP: Analysis in Progress
    Analysis Done: Analysis Done
    Development IP: Development in Progress
    Development Done: Development Done
    Test IP: Test in Progress
    Test Done: Test Done
    Done:
        - Closed
        - Done

# Map field names to additional attributes to extract
Attributes:
    Components: Component/s
    Priority: Priority
    Release: Fix version/s

If you are unfamiliar with YAML, remember that:

Comments start with #
Sections are defined with a name followed by a colon, and then an indented block underneath. Connection, Criteria, Workflow and Attributes area all sections in the example above.
Indentation has to use spaces, not tabs!
Single values can be set using Key: value pairs. For example, Project: ABC above sets the key Project to the value ABC.
Lists of values can be set by indenting a new block and placing a - in front of each list value. In the example above, the Issue types list contains the values Story and Defect.

The sections for Connection, Criteria and Workflow are required.

Under Conection, only Domain is required. If not specified, the script will prompt for both or either of username and password when run.

Under Criteria, all fields are technically optional, but you should specify at least some of them to avoid an unbounded query. Issue types and Valid resolutions can be set to either single values or lists.

Under Workflow, at least two steps are required. Specify the steps in order. You may either specify a single workflow value or a list (as shown for Done above), in which case multiple JIRA statuses will be collapsed into a single state for analytics purposes.

The file, and values for things like workflow statuses and attributes, are case insensitive.

When specifying attributes, use the name of the field (as rendered on screen in JIRA), not its id (as you might do in JQL), so e.g. use Component/s not components.

The attributes Type (issue type), Status and Resolution are always included.

When specifying fields like Component/s or Fix version/s that may have lists of values, only the first value set will be used.

Multiple queries

If it is difficult to construct a single set of criteria that returns all required issues, multiple Criteria sections can be wrapped into a Queries block, like so:

Queries:
    Attribute: Team
    Criteria:
        - Value: Team 1
          Project: ABC
          Issue types:
              - Story
              - Bug
          Valid resolutions:
              - Done
              - Closed
          JQL: Component = "Team 1"

        - Value: Team 2
          Project: ABC
          Issue types:
              - Story
              - Bug
          Valid resolutions:
              - Done
              - Closed
          JQL: Component = "Team 2"

In this example, the Component field in JIRA is being used to signify the team delivering the work, but may also be used for other things. Two JIRA queries will be run, corresponding to the two Criteria blocks.

In addition, a new column called Team will be added to the output, as specified by the Attribute field under Queries. For all items returned by the first query, the value will be Team 1 as per the Value field, and for all items returned by the second query, it will be Team 2.

Multi-valued fields

Some fields in JIRA can contain multiple values, e.g. fixVersion. By default, the extractor will use the first value in such a field if one is specified in the Attributes block. However, you may want to extract only specific values.

To do so, add a block like the following:

Attributes:
    Release: Fix version/s

Known values:
    Release:
        - "R01"
        - "R02"
        - "R03"

The extractor will pick the first "known value" found for the field. If none of the known values match, the cell will be empty.

Running

To produce the basic cycle time data, run jira-cycle-extract passing the name of the YAML configuration file and the name of the output CSV file:

$ jira-cycle-extract config.yaml data.csv

This will extract a CSV file called data.csv with cycle data based on the configuration in config.yaml, in a format compatible with the ActionableAgile toolset.

If you prefer Excel files for manual analysis:

$ jira-cycle-extract --format=xlsx config.yaml data.xlsx

If you prefer JSON:

$ jira-cycle-extract --format=json config.yaml data.json

The JSON format can be loaded by the Actionable Agile Analytics tool if you self-host it and the single-page HTML file for the AAA tool and the JSON file are accessible from the same web server, via a URL parameter:

http://myserver/analytics.html?url=data.json

You can specify a path or full URL, but due to same-origin request restrictions, your browser is unlikely to let you load anything not served from the same domain as the analytics web app itself.

Note: When the --format is set, it applies to all files written, not just the main cyle data file (see other options below). It is important to be consistent with the file extensions. In particular, if you are using the xlsx format you should also make sure all output files use a .xlsx extension.

There are lots more options. See:

$ jira-cycle-extract --help

Use the -v option to print more information during the extract process.

Use the -n option to limit the number of items fetched from JIRA, based on the most recently updated issues. This is useful for testing the configuration without waiting for long downloads:

$ jira-cycle-extract -v -n 10 config.yaml data.csv

To produce Cumulative Flow Diagram statistics, use the --cfd option:

$ jira-cycle-extract --cfd cfd.csv config.yaml data.csv

This will yield a cfd.csv file with one row for each date, one column for each step in the workflow, and a count of the number of issues in that workflow state on that day. To plot a CFD, chart this data as a (non-stacked) area chart. You should technically exclude the series in the first column if it represents the backlog!

To produce cycle time scatter plot statistics, use the --scatterplot option:

$ jira-cycle-extract --scatterplot scatterplot.csv config.yaml data.csv

This will yield a scatterplot.csv file with one row for each item that was completed (i.e. it reached the last workflow state), with columns giving the completion date and the number of days elapsed from the item entering the first active state (i.e. the second step in the workflow, on the basis that the first item represents a backlog or intake queue) to the item entering the completed state. These two columns can be plotted as an X/Y scatter plot. Further columns contain the dates of entry into each workflow state and the various issue metadata to allow further filtering.

To be able to easily draw a histogram of the cycle time values, use the --histogram option:

$ jira-cycle-extract --histogram histogram.csv config.yaml data.csv

This will yield a histogram.csv file with two columns: bin ranges and the number of items with cycle times falling within each bin. These can be charted as a column or bar chart.

To find out the 30th, 50th, 70th, 85th and 95th percentile cycle time values, pass the --percentiles option:

$ jira-cycle-extract --percentiles percentiles.csv config.yaml data.csv

To calculate different percentiles use the --quantiles option:

$ jira-cycle-extract --percentiles percentiles.csv --quantiles=0.3,0.5,0.8 config.yaml data.csv

Note that there should not be spaces between the commas!

To find out the daily throughput for the last 60 days, use the --throughput option:

$ jira-cycle-extract --throughput throughput.csv config.yaml data.csv

To use a different time window, e.g. the last 90 days:

$ jira-cycle-extract --throughput throughput.csv --throughput-window=90 config.yaml data.csv

The various options can be used in combination, and it is technically OK to skip the second positional (data.csv) parameter (in which case the file will not be written).

If you have charting dependencies installed (see above), there are various options available to allow you to draw charts, for example:

$ jira-cycle-extract --charts-scatterplot=scatterplot.png config.yaml data.csv

The available charts are:

--charts-scatterplot to draw a scatterplot of cycle times, with percentile lines
--charts-histogram to draw a histogram of cycle times, with percentile lines
--charts-cfd to draw a Cumulative Flow Diagram
--charts-throughput to draw a daily throughput bar chart
--charts-burnup to draw a simple burn-up chart (completed item count vs. time)
--charts-burnup-forecast to draw a burn-up chart with a Monte Carlo simulation showing paths towards a completion target. The completion target will by default be the number of items in the backlog, but can be set explicitly with the --charts-burnup-forecast-target options. The simluation by default uses 100 trials. The number of trials can be set with the --charts-burnup-forecast-trials option. You can set a deadline marker with the --charts-burnup-forecast-deadline option, which should be set to a date. If you also set --charts-burnup-forecast-deadline-confidence to a fraction (e.g. 0.85) it will be used to find a confidence interval in the simulation to which the deadline will be compared.
--charts-wip to draw a WIP boxplot showing min, max, median and mean WIP by week. By default, this will show the last 5 or 6 weeks' of data (depending on the weekday). You can change this with the --charts-wip-window parameter, set to a number of weeks.
--charts-ageing-wip to draw an ageing WIP chart: a scatter plot of current cycle time against state in the cycle, i.e. how items are trending towards completion.
--charts-net-flow to show a bar chart of the weekly net flow: departures - arrivals. By default, this will show the last 5 or 6 weeks' of data (depending on the weekday). You can change this with the --charts-net-flow-window parameter, set to a number of weeks.

Also note: all the --charts-* options have a corresponding --charts-*-title option that can be used to set a title for the chart.

Finally, to limit the date range of the data shown in the charts, you can use the options --charts-from and --charts-to to specify a starting and/or ending date (inclusive). Both are optional.

Troubleshooting

If Excel complains about a SYLK format error, ignore it. Click OK. See https://support.microsoft.com/en-us/kb/215591.
JIRA error messages may be printed out as HTML in the console. The error is in there somewhere, but may be difficult to see. Most likely, this is either an authentication failure (incorrect username/password or blocked account), or an error in the Criteria section resulting in invalid JQL.
If you aren't getting the issues you expected to see, use the -v option to see the JQL being sent to JIRA. Paste this into the JIRA issue filter search box ("Advanced mode") to see how JIRA evaluates it.
Old workflow states can still be part of an issue's history after a workflow has been modified. Use the -v option to find out about workflow states that haven't been mapped.
Excel sometimes picks funny formats for data in CSV files. Just set them to whatever makes sense.
If you are on a Mac and you get an error about Python not being installed as a framework, try to create a file ~/.matplotlib/matplotlibrc with the following contents:
```
backend : Agg
```
To install the charting dependencies on a Mac, you probably need to install a gfortran compiler for scipy. Use Homebrew (http://brew.sh) and install the gcc brew.

Ad-hoc analysis

Sometimes, you may want to perform more exploratory, ad-hoc analysis of the cycle data. jira-cycle-extract uses Python Pandas (http://pandas.pydata.org) to do most of its heavy lifting, and Pandas provides a rich environment for data science.

The Jupyter Notebook (http://jupyter.org) is a popular way to conduct interactive, ad-hoc analysis using Pandas (and more!).

If you have this running, here's an example of a notebook that uses jira-cycle-extract to query JIRA with a given YAML file configuration and makes the data available for further analysis:

import getpass
import datetime

import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import matplotlib as mpl
import seaborn as sns

from jira import JIRA
from jira_cycle_extract import cycletime, config

# Print charts in the notebook, using retina graphics
%matplotlib inline
%config InlineBackend.figure_format = 'retina'
sns.set_context("talk")

# Prompt for JIRA username, password and config file:
username = raw_input("Username:")
password = getpass.getpass("Password:")
config_filename = raw_input("Config file:")

# Parse options
options = {}
with open(config_filename) as config_file:
    options = config.config_to_options(config_file.read())

# Connect to JIRA
jira = JIRA(options={'server': options['connection']['domain']}, basic_auth=(username, password))

# Fetch issues and calculate cycle data as a Pandas DataFrame
q = cycletime.CycleTimeQueries(jira, **options['settings'])
cycle_data = q.cycle_data(verbose=False)

# Calculate other DataFrames for CFD, scatterplot, histogram, percentile and throughput data
cfd_data = q.cfd(cycle_data)
scatter_data = q.scatterplot(cycle_data)
histogram_data = q.histogram(cycle_data)
percentile_data = q.percentiles(cycle_data, percentiles=quantiles)
daily_throughput_data = q.throughput_data(cycle_data[cycle_data])

You can now do all kinds of analysis on the DataFrames (cycle_data, cfd_data and so on).

Changelog

0.10 - June 8 2016

Added title options for all charts
Added deadline option for burnup forecast chart

0.9 - May 31 2016

Add Docker documentation

0.8 - May 30 2016

Fixed a bug with calculating the CFD when statuses are skipped
Added --throughput output
Percentiles are now saved to file, not printed, when using --percentiles
Adding charting output (with optional dependencies - see above)

0.7 - January 22 2016

Add support for --format=json
Output all dates in ISO format (YYYY-MM-DD)

0.6 - January 20 2016

Add support for Queries and Known values.

0.5 - November 8 2015

When an issues moves between two JIRA states that are mapped to the same workflow step, record the earliest date, not the most recent
When an issue moves backwards in the flow as defined by the sequence of workflow steps, retain the earliest date the issue entered the given step (and erase any dates recorded for all subsequent steps)

0.4 - October 31 2015

Fix encoding errors when summary contains non-ASCII characters

0.3 - October 11 2015

Add proper support for --cfd, --scatterplot, --percentiles and --histogram
Fix some typing issues with the main cycle data extract.

0.2 - October 10 2015

Fix documentation errors

0.1 - October 10 2015

Initial release