I think this is going to be a rather complex debug. First, whoever is programming this function should really wrap their head around what the goal of this feature actually is: we are attempting to use old inputs (a previous xlsx) as the "old" reagent, or the "old stock solution" and remove the regions that these old reagents were able to access from the total region of the available concentration space.
The difference sampling is clearly not working either:
For example, in the case of a simple iodide perovskite experiment we can target the grey region as the full space. The blue region is what has already been sampled (old sampling, poor). The yellow region is the space that is accessible for all of the PbI2 containing compounds, while the yellow space highlighted by the black triangle is the region accessible by only 2 of the stock solutions.
For examples of the failings of the current multistock sampling see issue 96 folder in ESCALATE_development (here
A demonstration of a run accessing a related "yellow" region. LBL_Multistock_Edited.xlsx
A demonstration of a difference sampling run that should work but errors out: LBL_Multistock_Edited_subregion.xlsx
The fix for this is not straight forward. There really are two options: figure out what is actually needed to pull off the desired operation, or remove the "old_reagents" (terrible name) sections of the code entirely. Either works, but I think there is a benefit for being able to easily test / sample new regions with additional reagent definitions.
The best result would be to sample using new reagents for the spaces that were inaccessible using only the "old reagents". The generated runs will use the MINIMUM number of reagents possible to sample the new space and correctly prepare the run files.
Far above and beyond extension: Use the existing dataset of associated workflows (wf1.1 , or 3.0) and bound the area of explore space with a convex hull, figure out what regions haven't been explored and develop runs using the fewest number of reagents possible.
A few challenges
I think this is going to be a rather complex debug. First, whoever is programming this function should really wrap their head around what the goal of this feature actually is: we are attempting to use old inputs (a previous xlsx) as the "old" reagent, or the "old stock solution" and remove the regions that these old reagents were able to access from the total region of the available concentration space.
The difference sampling is clearly not working either:
For example, in the case of a simple iodide perovskite experiment we can target the grey region as the full space. The blue region is what has already been sampled (old sampling, poor). The yellow region is the space that is accessible for all of the PbI2 containing compounds, while the yellow space highlighted by the black triangle is the region accessible by only 2 of the stock solutions.
For examples of the failings of the current multistock sampling see issue 96 folder in ESCALATE_development (here
A demonstration of a run accessing a related "yellow" region. LBL_Multistock_Edited.xlsx
A demonstration of a difference sampling run that should work but errors out: LBL_Multistock_Edited_subregion.xlsx
The fix for this is not straight forward. There really are two options: figure out what is actually needed to pull off the desired operation, or remove the "old_reagents" (terrible name) sections of the code entirely. Either works, but I think there is a benefit for being able to easily test / sample new regions with additional reagent definitions.
The best result would be to sample using new reagents for the spaces that were inaccessible using only the "old reagents". The generated runs will use the MINIMUM number of reagents possible to sample the new space and correctly prepare the run files.
Far above and beyond extension: Use the existing dataset of associated workflows (wf1.1 , or 3.0) and bound the area of explore space with a convex hull, figure out what regions haven't been explored and develop runs using the fewest number of reagents possible.
A few challenges