Pandas dependency is to high

[zljubisic]

@fboerman I have seen closed discussion on #351.
What is said there is partially true.

I have checked the codebase (executed from entsoe-py root, don't use | head -n 10 to see all usages):

$ grep -irn --include=*.py "pd\." | head -n 10
entsoe/decorators.py:56:            df = pd.concat([df1, df2])
entsoe/decorators.py:82:            df = pd.concat(
entsoe/decorators.py:113:            not isinstance(start, pd.Timestamp)
entsoe/decorators.py:114:            or not isinstance(end, pd.Timestamp)
entsoe/decorators.py:130:                if func.__name__ != '_query_unavailability' and isinstance(frame.index, pd.DatetimeIndex):
entsoe/decorators.py:159:        df = pd.concat(frames, sort=True)
entsoe/decorators.py:187:        df = pd.concat(frames)
entsoe/entsoe.py:35:QUARTER_MTU_SDAC_GOLIVE =  pd.Timestamp('2025-10-01', tz='europe/amsterdam')
entsoe/entsoe.py:82:    def _base_request(self, params: Dict, start: pd.Timestamp,
entsoe/entsoe.py:83:                      end: pd.Timestamp) -> requests.Response:

Looks like entsoe-py uses very basic pandas functionality so there shouldn't be a problem to lower its version.

I have also played with dependencies a bit and I can say that this requirements.txt works with project's tests.

$ cat requirements.txt
requests
pytz
beautifulsoup4>=4.11.1
numpy>=1.26.4
pandas>=1.5.3

I have created a new (conda) environment, used pip install -r requirements, set entsoe uname/pass, installed pytest and dotenv and run pytest with no issues.

I don't know if tests (there are only two of them in tests directory) are covering all cases, but it seams pandas/numpy could be at much lower versions.

All dependencies should be at the lowest supported versions with >= to allow any greater version.

[fboerman]

Hi! The tests only cover the new file subpackage. Other functions require higher pandas version mostly due to changing frequency strings. I guess this is a good reminder I should work on the PR that adds more tests but I really didn't have the time for such a big thing. I am not lowering the version until the tests are ready because I know it caused crashes in some functions. ⁣Get BlueMail for Android ​

…

On 19 Sept 2025, 12:29, at 12:29, zljubisic ***@***.***> wrote: zljubisic created an issue (EnergieID/entsoe-py#437) @fboerman I have seen closed discussion on #351. What is said there is partially true. I have checked the codebase (executed from `entsoe-py` root, don't use `| head -n 10` to see all usages): ```shell $ grep -irn --include=*.py "pd\." | head -n 10 entsoe/decorators.py:56: df = pd.concat([df1, df2]) entsoe/decorators.py:82: df = pd.concat( entsoe/decorators.py:113: not isinstance(start, pd.Timestamp) entsoe/decorators.py:114: or not isinstance(end, pd.Timestamp) entsoe/decorators.py:130: if func.__name__ != '_query_unavailability' and isinstance(frame.index, pd.DatetimeIndex): entsoe/decorators.py:159: df = pd.concat(frames, sort=True) entsoe/decorators.py:187: df = pd.concat(frames) entsoe/entsoe.py:35:QUARTER_MTU_SDAC_GOLIVE = pd.Timestamp('2025-10-01', tz='europe/amsterdam') entsoe/entsoe.py:82: def _base_request(self, params: Dict, start: pd.Timestamp, entsoe/entsoe.py:83: end: pd.Timestamp) -> requests.Response: ``` Looks like `entsoe-py` uses very basic pandas functionality so there shouldn't be a problem to lower its version. I have also played with dependencies a bit and I can say that this `requirements.txt` works with project's tests. ```shell $ cat requirements.txt requests pytz beautifulsoup4>=4.11.1 numpy>=1.26.4 pandas>=1.5.3 ``` I have created a new (conda) environment, used `pip install -r requirements`, set entsoe uname/pass, installed `pytest` and `dotenv` and run `pytest` with no issues. I don't know if tests (there are only two of them in tests directory) are covering all cases, but it seams `pandas/numpy` could be at much lower versions. All dependencies should be at the lowest supported versions with `>=` to allow any greater version. -- Reply to this email directly or view it on GitHub: #437 You are receiving this because you were mentioned. Message ID: ***@***.***>

[zljubisic]

I am really curious which pandas method would cause problems here.
If you can show me a test case that works with pandas 2.x and doesn't work with pandas 1.x I can probably help with solution.

Tests shouldn't be a problem today. With a little help of chatgpt they can be created easily.
Problem is that you must know what you are testing and where problems are expected.