Relaxed on requirement for parameters.txt

[alifbe]

In some cases, we run prediction using restart file from other ensembles. For example in Drogon, we have a prediction run that is based on iter-3 of automatic history matching (AHM) model. For these cases, there won't be parameters.txt in the RUNPATH. FMU Ensemble will not read the data from those cases and causing crash in Webviz plugin StructuralUncertainty.

I have tried to put an empty parameters.txt and the plugin seems to work fine.

So, I am suggesting to relaxed the requirement for parameters.txt in this code

fmu-ensemble/src/fmu/ensemble/ensemble.py

Lines 410 to 422 in d63a380

	def parameters(self):
	"""Build a dataframe of the information in each
	realizations parameters.txt.
	If no realizations have the file, an empty dataframe is returned.
	Returns:
	pd.DataFrame
	"""
	try:
	return self.load_txt("parameters.txt")
	except KeyError:
	return pd.DataFrame()

[asnyv]

Sorry that you haven't gotten any response on this.

From the code you refer to, it seems like an empty dataframe is returned, and not that it crashes. That makes sense to me: no data -> blank output? the alternative is to maybe still add the realization column and nothing else.
But based on the attached code, this seems to me like a problem that likely needs to be resolved in the webviz plugin you refer to

[asnyv]

But: Even for restarts / prediction runs it is generally recommended to add uncertainty.

1. It is often useful to transfer the parameters from history into prediction to see if you can get some insight into which AHM parameters that affect the predictions, considering predictions are what really matters (the history is water under the bridge)
2. You generally have other factors that are uncertain for the predictions, e.g. minimum flow in wells and pipelines, risk factors for sand / scale issues or other things that affect productivity, capacity limitations on water handling, timing of future projects, productivity of new wells and so on.

[alifbe]

Ideally, it will be great if it loads the result per realization into the data frame instead of returning empty data frame because these data are needed for the plugin.

fmu-ensemble/src/fmu/ensemble/ensemble.py

Lines 486 to 520 in d63a380

	def load_file(self, localpath, fformat, convert_numeric=False, force_reread=False):
	"""Function for calling load_file() in every realization
	This function may utilize multithreading.
	Args:
	localpath (str): path to the text file, relative to each realization
	fformat (str): string identifying the file format. Supports 'txt'
	and 'csv'.
	convert_numeric (boolean): If set to True, numerical columns
	will be searched for and have their dtype set
	to integers or floats. If scalars, only numerical
	data will be loaded.
	force_reread (boolean): Force reread from file system. If
	False, repeated calls to this function will
	returned cached results.
	Returns:
	pd.Dataframe: with loaded data aggregated. Column 'REAL'
	distuinguishes each realizations data.
	"""
	for index, realization in self.realizations.items():
	try:
	realization.load_file(localpath, fformat, convert_numeric, force_reread)
	except ValueError as exc:
	# This would at least occur for unsupported fileformat,
	# and that we should not skip.
	logger.critical("load_file() failed in realization %d", index)
	raise ValueError from exc
	except IOError:
	# At ensemble level, we allow files to be missing in
	# some realizations
	logger.warning("Could not read %s for realization %d", localpath, index)
	if self.get_df(localpath).empty:
	raise ValueError("No ensemble data found for {}".format(localpath))
	return self.get_df(localpath)

In fmu-drogon, we have created a forward model that transfer the parameters.txt from history (equinor/fmu-drogon@a7aa6d5), but I just think it will be great that we can still support the runs that are already performed (without this new forward model)