summary refs log tree commit diff
path: root/docs/development/dependencies.md
blob: 236856a6b047a8ca1c2a77f84463f311e911aff7 (plain) (blame)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
# Managing dependencies with Poetry

This is a quick cheat sheet for developers on how to use [`poetry`](https://python-poetry.org/).

# Background

Synapse uses a variety of third-party Python packages to function as a homeserver.
Some of these are direct dependencies, listed in `pyproject.toml` under the
`[tool.poetry.dependencies]` section. The rest are transitive dependencies (the
things that our direct dependencies themselves depend on, and so on recursively.)

We maintain a locked list of all our dependencies (transitive included) so that
we can track exactly which version of each dependency appears in a given release.
See [here](https://github.com/matrix-org/synapse/issues/11537#issue-1074469665)
for discussion of why we wanted this for Synapse. We chose to use
[`poetry`](https://python-poetry.org/) to manage this locked list; see
[this comment](https://github.com/matrix-org/synapse/issues/11537#issuecomment-1015975819)
for the reasoning.

The locked dependencies get included in our "self-contained" releases: namely,
our docker images and our debian packages. We also use the locked dependencies
in development and our continuous integration.

Separately, our "broad" dependencies—the version ranges specified in
`pyproject.toml`—are included as metadata in our "sdists" and "wheels" [uploaded
to PyPI](https://pypi.org/project/matrix-synapse). Installing from PyPI or from
the Synapse source tree directly will _not_ use the locked dependencies; instead,
they'll pull in the latest version of each package available at install time.

## Example dependency

An example may help. We have a broad dependency on
[`phonenumbers`](https://pypi.org/project/phonenumbers/), as declared in
this snippet from pyproject.toml [as of Synapse 1.57](
https://github.com/matrix-org/synapse/blob/release-v1.57/pyproject.toml#L133
):

```toml
[tool.poetry.dependencies]
# ...
phonenumbers = ">=8.2.0"
```

In our lockfile this is
[pinned]( https://github.com/matrix-org/synapse/blob/dfc7646504cef3e4ff396c36089e1c6f1b1634de/poetry.lock#L679-L685)
to version 8.12.44, even though
[newer versions are available](https://pypi.org/project/phonenumbers/#history).

```toml
[[package]]
name = "phonenumbers"
version = "8.12.44"
description = "Python version of Google's common library for parsing, formatting, storing and validating international phone numbers."
category = "main"
optional = false
python-versions = "*"
```

The lockfile also includes a
[cryptographic checksum](https://github.com/matrix-org/synapse/blob/release-v1.57/poetry.lock#L2178-L2181)
of the sdists and wheels provided for this version of `phonenumbers`.

```toml
[metadata.files]
# ...
phonenumbers = [
    {file = "phonenumbers-8.12.44-py2.py3-none-any.whl", hash = "sha256:cc1299cf37b309ecab6214297663ab86cb3d64ae37fd5b88e904fe7983a874a6"},
    {file = "phonenumbers-8.12.44.tar.gz", hash = "sha256:26cfd0257d1704fe2f88caff2caabb70d16a877b1e65b6aae51f9fbbe10aa8ce"},
]
```

We can see this pinned version inside the docker image for that release:

```
$ docker pull matrixdotorg/synapse:v1.57.0
...
$ docker run --entrypoint pip matrixdotorg/synapse:v1.57.0 show phonenumbers
Name: phonenumbers
Version: 8.12.44
Summary: Python version of Google's common library for parsing, formatting, storing and validating international phone numbers.
Home-page: https://github.com/daviddrysdale/python-phonenumbers
Author: David Drysdale
Author-email: dmd@lurklurk.org
License: Apache License 2.0
Location: /usr/local/lib/python3.9/site-packages
Requires:
Required-by: matrix-synapse
```

Whereas the wheel metadata just contains the broad dependencies:

```
$ cd /tmp
$ wget https://files.pythonhosted.org/packages/ca/5e/d722d572cc5b3092402b783d6b7185901b444427633bd8a6b00ea0dd41b7/matrix_synapse-1.57.0rc1-py3-none-any.whl
...
$ unzip -c matrix_synapse-1.57.0rc1-py3-none-any.whl matrix_synapse-1.57.0rc1.dist-info/METADATA | grep phonenumbers
Requires-Dist: phonenumbers (>=8.2.0)
```

# Tooling recommendation: direnv

[`direnv`](https://direnv.net/) is a tool for activating environments in your
shell inside a given directory. Its support for poetry is unofficial (a
community wiki recipe only), but works solidly in our experience. We thoroughly
recommend it for daily use. To use it:

1. [Install `direnv`](https://direnv.net/docs/installation.html) - it's likely
   packaged for your system already.
2. Teach direnv about poetry. The [shell config here](https://github.com/direnv/direnv/wiki/Python#poetry)
   needs to be added to `~/.config/direnv/direnvrc` (or more generally `$XDG_CONFIG_HOME/direnv/direnvrc`).
3. Mark the synapse checkout as a poetry project: `echo layout poetry > .envrc`.
4. Convince yourself that you trust this `.envrc` configuration and project.
   Then formally confirm this to `direnv` by running `direnv allow`.

Then whenever you navigate to the synapse checkout, you should be able to run
e.g. `mypy` instead of `poetry run mypy`; `python` instead of
`poetry run python`; and your shell commands will automatically run in the
context of poetry's venv, without having to run `poetry shell` beforehand.


# How do I...

## ...reset my venv to the locked environment?

```shell
poetry install --extras all --remove-untracked
```

## ...run a command in the `poetry` virtualenv?

Use `poetry run cmd args` when you need the python virtualenv context.
To avoid typing `poetry run` all the time, you can run  `poetry shell`
to start a new shell in the poetry virtualenv context. Within `poetry shell`,
`python`, `pip`, `mypy`, `trial`, etc. are all run inside the project virtualenv
and isolated from the rest o the system.

Roughly speaking, the translation from a traditional virtualenv is:
- `env/bin/activate` -> `poetry shell`, and
- `deactivate` -> close the terminal (Ctrl-D, `exit`, etc.)

See also the direnv recommendation above, which makes `poetry run` and
`poetry shell` unnecessary.


## ...inspect the `poetry` virtualenv?

Some suggestions:

```shell
# Current env only
poetry env info
# All envs: this allows you to have e.g. a poetry managed venv for Python 3.7,
# and another for Python 3.10.
poetry env list --full-path
poetry run pip list
```

Note that `poetry show` describes the abstract *lock file* rather than your
on-disk environment. With that said, `poetry show --tree` can sometimes be
useful.


## ...add a new dependency?

Either:
- manually update `pyproject.toml`; then `poetry lock --no-update`; or else
- `poetry add packagename`. See `poetry add --help`; note the `--dev`,
  `--extras` and `--optional` flags in particular.
  - **NB**: this specifies the new package with a version given by a "caret bound". This won't get forced to its lowest version in the old deps CI job: see [this TODO](https://github.com/matrix-org/synapse/blob/4e1374373857f2f7a911a31c50476342d9070681/.ci/scripts/test_old_deps.sh#L35-L39).

Include the updated `pyproject.toml` and `poetry.lock` files in your commit.

## ...remove a dependency?

This is not done often and is untested, but

```shell
poetry remove packagename
```

ought to do the trick. Alternatively, manually update `pyproject.toml` and
`poetry lock --no-update`. Include the updated `pyproject.toml` and poetry.lock`
files in your commit.

## ...update the version range for an existing dependency?

Best done by manually editing `pyproject.toml`, then `poetry lock --no-update`.
Include the updated `pyproject.toml` and `poetry.lock` in your commit.

## ...update a dependency in the locked environment?

Use

```shell
poetry update packagename
```

to use the latest version of `packagename` in the locked environment, without
affecting the broad dependencies listed in the wheel.

There doesn't seem to be a way to do this whilst locking a _specific_ version of
`packagename`. We can workaround this (crudely) as follows:

```shell
poetry add packagename==1.2.3
# This should update pyproject.lock.

# Now undo the changes to pyproject.toml. For example
# git restore pyproject.toml

# Get poetry to recompute the content-hash of pyproject.toml without changing
# the locked package versions.
poetry lock --no-update
```

Either way, include the updated `poetry.lock` file in your commit.

## ...export a `requirements.txt` file?

```shell
poetry export --extras all
```

Be wary of bugs in `poetry export` and `pip install -r requirements.txt`.

Note: `poetry export` will be made a plugin in Poetry 1.2. Additional config may
be required.

## ...build a test wheel?

I usually use

```shell
poetry run pip install build && poetry run python -m build
```

because [`build`](https://github.com/pypa/build) is a standardish tool which
doesn't require poetry. (It's what we use in CI too). However, you could try
`poetry build` too.


# Troubleshooting

## Check the version of poetry with `poetry --version`.

At the time of writing, the 1.2 series is beta only. We have seen some examples
where the lockfiles generated by 1.2 prereleasese aren't interpreted correctly
by poetry 1.1.x. For now, use poetry 1.1.14, which includes a critical
[change](https://github.com/python-poetry/poetry/pull/5973) needed to remain
[compatible with PyPI](https://github.com/pypi/warehouse/pull/11775).

It can also be useful to check the version of `poetry-core` in use. If you've
installed `poetry` with `pipx`, try `pipx runpip poetry list | grep poetry-core`.

## Clear caches: `poetry cache clear --all pypi`.

Poetry caches a bunch of information about packages that isn't readily available
from PyPI. (This is what makes poetry seem slow when doing the first
`poetry install`.) Try `poetry cache list` and `poetry cache clear --all
<name of cache>` to see if that fixes things.

## Try `--verbose` or `--dry-run` arguments.

Sometimes useful to see what poetry's internal logic is.