Deprecate Python 2.6 and 3.3 before removal

[hugovk]

It's a good idea to deprecate before removal (#166), to give downstream dependencies a heads up.

How would you like to deprecate?

Here's some ideas. Let me know if you want any of them removing or changing. I assumed JMESPath 1.0 would be the version they'll be dropped in, but that of course can be changed.

Warn when importing

• Pro: deprecation is visible
• Con: maybe too visible

Warn when installing

• Pro: gets the message across, not too often
• Con: maybe you don't want any warning

Document in CHANGELOG

• The minimum needed

Also added python_requires for the current versions (2.6+, 3.3+). This would be useful whatever is decided.

[jeking3]

Given folks can lock to a previous release, is deprecation necessary?

[hugovk]

See #166 (comment):

I would love to drop support for python2.6, but this would affect many downstream dependencies that use jmespath. If we're going to do this, we'd need to have a deprecation period first to give people a heads up.

[jeking3]

I'm pretty sure that it would not, since folks downstream with well-designed requirements files / setup.py / whatever they use to express dependencies will have already locked their builds to a specific major or major.minor of jmespath which would keep them compatible with python 2.6 until they can move forward.

[jamesls]

Thanks for starting this process!

[codecov-io]

Codecov Report

Merging #175 into develop will increase coverage by <.01%.
The diff coverage is 100%.

@@            Coverage Diff             @@
##           develop    #175      +/-   ##
==========================================
+ Coverage    98.19%   98.2%   +<.01%     
==========================================
  Files           14      14              
  Lines         1441    1446       +5     
==========================================
+ Hits          1415    1420       +5     
  Misses          26      26

Impacted Files	Coverage Δ
jmespath/__init__.py	100% <100%> (ø)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 0d5762b...a5c9b22. Read the comment docs.

[hugovk]

Updated to resolve conflicts!

---

I'd suggest moving forward with this process, as you can see Python 2.6 and 3.3 are not available on Travis CI:

$ curl -sSf -o python-2.6.tar.bz2 ${archive_url}
curl: (22) The requested URL returned error: 404 Not Found
Unable to download 2.6 archive. The archive may not exist. Please consider a different version.

$ curl -sSf -o python-3.3.tar.bz2 ${archive_url}
curl: (22) The requested URL returned error: 404 Not Found
Unable to download 3.3 archive. The archive may not exist. Please consider a different version.

https://travis-ci.org/jmespath/jmespath.py/builds/569774974

---

Also, Python 3.4 is also now EOL and no longer receiving security updates (or any updates) from the core Python team.

Version	Release date	Supported until
2.7	2010-07-03	2020-01-01
3.4	2014-03-16	2019-03-16

I suggest deprecating/removing that as well.

[and-semakin]

Wow, Python 2.6 has reached its EOL in October, 2013. More then six years ago! I hope that all folks who uses it are already well prepared. I think that it (at least 2.6) can be dropped even without deprecation warning.

[jamesls]

I know it's been a while, but I want to follow up on this. Here's my plan.

I think it's fine to move fairly quickly on deprecating 2.6/3.3, especially given some of the largest downstream dependencies have dropped 2.6/3.3 recently. Initially I was surprised at the large number of 2.6 downloads, but in looking into it more, it seems to closely match the 2.6 numbers of the aws-cli/botocore which has already dropped 26/33 support.

| category | percent | downloads  |
|----------|--------:|-----------:|
| 2.7      |  51.87% | 22,981,642 |
| 3.7      |  16.73% |  7,411,660 |
| 3.6      |  13.75% |  6,092,117 |
| 2.6      |   6.98% |  3,090,928 |
| 3.5      |   5.05% |  2,237,316 |
| 3.8      |   3.34% |  1,481,817 |
| 3.4      |   1.77% |    785,991 |
| null     |   0.50% |    222,783 |
| 3.9      |   0.00% |        482 |
| 3.3      |   0.00% |        222 |
| 3.2      |   0.00% |          4 |
| Total    |         | 44,304,962 |

Date range: 2020-01-01 - 2020-01-31

So here's my plan:

• I'm going to get a 0.9.5 release out that adds support for py38
• I've created (Update travis to pass on all supported python versions #194) that fixes travis for 2.6. I'll merge that and then rebase this PR with those changes.
• I'll cut 0.10.0 that officially deprecates 2.6/3.3.
• The next version, 0.11.0, will remove support for 2.6/3.3. I'll give this about a month, so I'll plan on a release March 30th.

Let me know if there's any questions/concerns.

[jamesls]

Merged via #195