Pypi fix (#109)

* Dependencies from git sources moved from `setup.py` to `requirements-dev.txt`.
- README.md updated.

* Package source added to requirement message helper function, so that packages available from git source can now be inherently added.

* CI updated with requirements-dev.txt.

* Pip installations orders changed.
- Typo fix in utils.py

* Package `click` is fixed at version 8.0.4.

.github/workflows/ci.yml

@@ -54,6 +54,7 @@ jobs:
       - name: Install dependencies
         run: >
           pip install -e .[dev]
+          pip install -r requirements-dev.txt
 
       - name: Lint with flake8 and black
         run: |

README.md

@@ -88,6 +88,11 @@ or build from source,
     cd jury
     python setup.py install
 
+**NOTE:** There may be malfunctions of some metrics depending on `sacrebleu` package on Windows machines which is 
+mainly due to the package `pywin32`. For this, we fixed pywin32 version on our setup config for Windows platforms. 
+However, if pywin32 causes trouble in your environment we strongly recommend using `conda` manager install the package 
+as `conda install pywin32`.
+
 ## <div align="center"> Usage </div>
 
 ### API Usage
@@ -227,6 +232,14 @@ PRs are welcomed as always :)
     cd jury
     pip install -e .[dev]
 
+Also, you need to install the packages which are available through a git source separately with the following command. 
+For the folks who are curious about "why?"; a short explaination is that PYPI does not allow indexing a package which 
+are directly dependent on non-pypi packages due to security reasons. The file `requirements-dev.txt` includes packages 
+which are currently only available through a git source, or they are PYPI packages with no recent release or 
+incompatible with Jury, so that they are added as git sources or pointing to specific commits.
+
+    pip install -r requirements-dev.txt
+
 ### Tests
 
 To tests simply run.

jury/metrics/_core/utils.py

@@ -25,18 +25,19 @@ def __init__(self, path: str, task: str):
         super(TaskNotAvailable, self).__init__(message)
 
 
-def requirement_message(path: str, package_name: str) -> str:
+def requirement_message(path: str, package_name: str, package_source: str = None) -> str:
+    package_source = package_source or package_name
     return (
         f"In order to use metric '{path}', '{package_name}' is required. "
-        f"You can install the package by `pip install {package_name}`."
+        f"You can install the package by `pip install {package_source}`."
     )
 
 
 def download(source: str, destination: str, overwrite: bool = False, warn: bool = False) -> None:
     if os.path.exists(destination) and not overwrite:
         if warn:
             warnings.warn(
-                f"Path {destination} already exists, not overwriting. To overwrite, speficy " f"'overwrite' parameter."
+                f"Path {destination} already exists, not overwriting. To overwrite, speficy 'overwrite' parameter."
             )
         return
     r = requests.get(source, allow_redirects=True)

jury/metrics/bleurt/bleurt_for_language_generation.py

@@ -24,7 +24,7 @@
 import datasets
 
 from jury.metrics import LanguageGenerationInstance, MetricForLanguageGeneration
-from jury.metrics._core.utils import PackagePlaceholder
+from jury.metrics._core.utils import PackagePlaceholder, requirement_message
 
 # `import bleurt` placeholder
 bleurt = PackagePlaceholder(version="1.2.2")
@@ -122,8 +122,11 @@ def _download_and_prepare(self, dl_manager):
             from bleurt import score
         except ModuleNotFoundError:
             raise ModuleNotFoundError(
-                f"In order to use metric BLEURT, 'bleurt' is required. "
-                f"You can install the package by `bleurt @ pip install git+https://github.com/google-research/bleurt.git`."
+                requirement_message(
+                    path="bleurt",
+                    package_name="bleurt",
+                    package_source="git+https://github.com/devrimcavusoglu/bleurt.git",
+                )
             )
 
         # check that config name specifies a valid BLEURT model

requirements-dev.txt

@@ -0,0 +1,7 @@
+# Jury metrics
+# Falling back to fork for Windows related issue: https://github.com/google-research/bleurt/issues/49
+bleurt @ git+https://github.com/devrimcavusoglu/bleurt.git
+unbabel-comet @ git+https://github.com/Unbabel/COMET.git@c772b679e20725e6cc79b2107d50594f9ea7a4ae
+
+# for datasets metric test
+math_equivalence @ git+https://github.com/hendrycks/math.git

requirements.txt

@@ -1,3 +1,4 @@
+click==8.0.4
 datasets>=2.0.0
 fire>=0.4.0
 nltk>=3.6.6,<3.7.1

setup.py

@@ -28,7 +28,7 @@ def get_version():
 def add_pywin(reqs: List[str]) -> None:
     if platform.system() == "Windows":
         # Latest PyWin32 build (301) fails, required for sacrebleu
-        ext_package = ["pywin32==228"]
+        ext_package = ["pywin32==302"]
     else:
         ext_package = []
     reqs.extend(ext_package)
@@ -42,20 +42,18 @@ def add_pywin(reqs: List[str]) -> None:
     "pytest>=7.0.1",
     "pytest-cov>=3.0.0",
     "pytest-timeout>=2.1.0",
-    "math_equivalence @ git+https://github.com/hendrycks/math.git",  # for datasets test metric
 ]
 
 _PRISM_REQUIREMENTS = ["fairseq==0.9.0", "validators"]
 
 _METRIC_REQUIREMENTS = [
     "sacrebleu>=2.0.0",
     "bert_score==0.3.11",
-    "bleurt @ git+https://github.com/google-research/bleurt.git",
     "jiwer>=2.3.0",
     "seqeval==1.2.2",
     "sentencepiece==0.1.96",
-    "unbabel-comet @ git+https://github.com/Unbabel/COMET.git@c772b679e20725e6cc79b2107d50594f9ea7a4ae",
 ]
+
 _METRIC_REQUIREMENTS.extend(_PRISM_REQUIREMENTS)
 add_pywin(_METRIC_REQUIREMENTS)