simonensemble / porematmod.jl Goto Github PK

moiety should have kwargs for setting bonding rules or covalency parameters

should probably be added to module tests

needs to be fully explored in an example notebook

wasn't working last time I tried

examples page should be links (w/ illustrations and ultra-short blurbs) to Pluto notebooks of example and tutorial code

need more tests

want a function that takes q_o and q_o \ q_! and returns masked q

for simple replacements, easy

for more complex ones, e.g. disorder correction, would need to re-use the OP alignment testing code

not hard to do ad-hoc, but would be nice to have a real option built in for this

• export to html
• link in manual

would be nice to have better API for analyzing search results based on symmetry

e.g., searching for p-phenylene in IRMOF-1 gives 96 isom's at 24 loc's. Naïve assumption: there are a huge number of different structures we can make with just one replacement. But, no matter which of the 96 isom's is chosen for an initial replacement, the result is identical by symmetry. If you want to make a second replacement from there, the naïve assumption would be that there are now 23 loc's, each w/ 4 uniquely-oriented isom's. There are actually far fewer, because again, symmetry makes many of these equivalent choices.

properly handling this could be a real rabbit hole, and will definitely involve group theory and a nice dose of linear algebra

currently, must do moiety(nothing)

make substructure_replace and sugared form(s) accept simply nothing

child.name is always "new_xtal"

• add option to set name
• generate better automatic name

:cross_pb and :distance are being dropped, causing downstream problems with rendering in MOFunGO

only runs as MOFun.set_path_to_crystals

all cross-PB replacements are getting flipped around! example2.jl shows this clearly, as all cross-pb linkers now have C-O bonds across the boundary (so linker is backwards!)

get rid of the dependency on nothing.xyz for creating the "nothing" moiety

overall, missing clear documentation.

edit to make variable nomenclature consistent with lit/pub

this actually seems to be an issue with Crystal slicing in Xtals

drop_cross_pb_bonds! fails on sliced Crystal with KeyError: key :cross_boundary not found

add examples for creating missing-linker and missing-node crystal defects in UiO-66

not running

99% of search runtime is spent on prune!. Up to half of that is spent on calls to candidate_list.

Refactoring to use a Boolean matrix on the DFS stack for the candidate list could make searches much faster!

• register package
• remove pre-registration hacks

need an illustration for:

• each use case example
• Ullmann, find, OP, find-replace algos
• each input file type
• each replacement mode
• symmetry/rotation (BDC w/ color-coded atoms)

• mirror example 1 in others
• list uses cases like in paper for index.md
• start.md: don't need Xtals
• find.md: graphic from paper
• render Pluto notebooks as HTML
• make sure docs are building via Travis
• make sure Pluto HTML is auto-updated
• add notebook HTMLs to sidebar
• verify README link to docs
• verify all asset links in docs
• remove redundant examples folder
• check Pluto notebooks in doc tests?
• update examples.zip on docs build?
• more pictures in the docs
• format nav bar w/ headings for manual and examples?
• re-export SIFSIX structure to make the Crystal command to load it simpler

candidate_list() is inefficient due to branching. refactor to candidate_in_graph() and candidate_in_subgraph()

also need to make sure Travis will run those

need to implement this. how to determine exact manner of replacement?

usage case 1: provide function with user-specified replacement scheme

usage case 2: attempt to automatically infer a good replacement, probably via Ullmann

The existing syntactic sugar for substructure replacements may be a bit opaque, despite its beauty:

(q => r) in p

The overloading of the in operator is the issue (b/c q in p functions as the search command)

May be better to refactor as analogue of Base.replace:

replace(p, q => r)

• make repo public
• register package
• remove pre-registration hacks from examples and PMMG

• in source
• in docs
• in repo title

this is probably not desirable exactly, but also doesn't affect much... probably should change at some point

• default replacement behavior: optimal ori @ all loc
• kwarg for default: all_loc
• assert minimum 3 atoms in OP
• refactor isoms as Dict

the two definitions of replace in find_replace.jl generate a warning during precompilation that the second overwrites the first, but that is not true. why does this happen, and how can it be prevented?

• make docs consistent with code changes
• make docs consistent with paper language
• use paper figure code for doc examples

innermost loop code of refine_M!() causes repeated calls to neighbors() which can be avoided by moving the call to neighbors() up one loop level.

should make a video and post on YouTube showing how to perform a complete example find-replace, including data prep

need to implement possibility matrix sorting based on subgraph node degree, followed by un-sorting of result matrices

need multivariate MOF example as one of the advanced use case tutorials

multiple isomorphisms between the search moiety mask and replacement moiety can cause a problem with alignment. current solution is for user to provide "better" inputs... would be nice to make the software more intelligent here.

add link to paper pre-print once it's on ArXiv

update to final publication

not running

stack overflow occurs during search for p-phenylene in Ti-MIL-125. unknown if this is a program flow error or memory limitation issue.

link debug output formatting to stack depth for help in understanding algo flow?

Currently, OP for aligning r_c to p' is done in two steps, allowing greater error propagation (especially if p', q_c, and r_c are not the exact same geometry). This is done because the mapping \phi needs to be chosen with the assistance of OP.

Refactor so that once \phi is selected, the final OP is performed using \phi and \theta together.

would be super cool to be able to direct find-replace operations in MOFunGO with interactive tagging of search moiety atoms

(don't assume .xyz, even though it's the only supported type right now)

make the docs more cohesive with the IRMOF-1 functionalization story

• avoid using MultiSelect (unstable behavior)
• figure out what's wrong w/ drop_cross_pb_bonds!

there are cases where it would be beneficial to allow only isomorphisms between q and induced subgraphs of p, e.g. searching for n-propyl in the presence of cyclopropyl

add a flag for this in substructure_search