ferchault / iago Goto Github PK

Steps:

1. Generate sample output and add it to the repository.
2. Write Reader class to extract and convert data.

Introduce a new flag to plane fitting: raise_error. If True, undetermined planes due to missing atoms or collinearity raise an error. Default False.

Currently, we can understand CP2K log and input files. We should add the following codes: namd2, AMBER, gromacs.

Since the data structure in the database is already prepared this consists of the following steps:

1. For each code, implement a "Reader" class that takes the input file and (optionally) the output file. Parse these text-based files using regular expressions or similar frameworks. Test files and guidance can be provided.
2. Prepare unit tests that work on small test input and log files and compare the actual and expected output.

Optionally, suggest additional data special to the codes to be included in the database.

CP2K may print the coordinates in XYZ format which does not contain information on the box size.
With ensembles NVE/NVT: this information are present in the input file.
With ensembles NPT: in another file.

Now: use DCD format only.

Steps:

1. Change JSON to HDF5 database.
2. Use blaze for access. If SSH, try to use blaze server

Within the UCL network, we can connect with ssh the UCL machines without using the socrates gateway.

It could be useful to connect the UCL machines from outside.

CP2K may be used with more than PSF topology files (e.g. one PSF per different kind of molecules). The code is not ready for that.
Now: modify the CP2K input to print a unique full-system PSF using:
https://manual.cp2k.org/cp2k-4_1-branch/CP2K_INPUT/FORCE_EVAL/SUBSYS/TOPOLOGY/DUMP_PSF.html

For analysing atomistic trajectories, the gromacs code has well-tested and fast C++ implementations. We should leverage this by converting our trajectory data to their format (there are open-source libraries for that) and then feed it to their toolchain. Finally, the output (plain-text) has to be parsed back for storage in the database.

Steps:

1. Understand the gromacs input requirements (guidance available).
2. Write a python code to convert the file formats (using the MDAnalysis library).
3. Write a wrapper for the various gromacs tools exposing their command line options in a python interface.
4. Convert their output to the iago internal database structure.

Create integration tests for file access using the unit test framework.

The documentation-Get Started need updated:
import iago
lg = iago.get_location_group()
lg.get_bucket_list()
db = lg.fetch_database('ubqTUTORIAL')

If do not have get_bucket_list() before fetch_database:
import iago
lg = iago.get_location_group()
\#lg.get_bucket_list()
db = lg.fetch_database('ubqTUTORIAL')
The error messages are the following:

AttributeError Traceback (most recent call last)
in ()
2 lg = iago.get_location_group()
3 #lg.get_bucket_list()
----> 4 db = lg.fetch_database('ubqTUTORIAL')

/scratch/xjiang/anaconda2/lib/python2.7/site-packages/iago/LocationProvider.pyc in fetch_database(self, bucket)
121
122 def fetch_database(self, bucket):
--> 123 row = self._get_bucket_from_name_or_id(bucket)
124 format = None
125 ephemeral = False

/scratch/xjiang/anaconda2/lib/python2.7/site-packages/iago/LocationProvider.pyc in _get_bucket_from_name_or_id(self, bucket)
108
109 def _get_bucket_from_name_or_id(self, bucket):
--> 110 if bucket in self._buckets.id.values:
111 rows = self._buckets[self._buckets.id == bucket]
112 elif bucket in self._buckets.name.values:

AttributeError: 'NoneType' object has no attribute 'id'

Expose energy-like data to the user.

When investigating trajectories, it is often helpful to get a quick visualisation of parts of the system.

We should create a jupyter notebook widget that can show the 3D coordinates of atoms in space either as 3d rotatable frame or as projection on an arbitrary rotatable 2d plane. We should be able to update the geometry and show the video when the user seeks through the molecular dynamics data.

It is crucial that this visualisation is fast and interactive. We should support major browsers, but may focus on OpenGL if required.

Class method iago.Analyser.dynamic_distance() may not properly support non-rectangular unit cells. Fix code and add test case.

On the way towards windows support, alternatives for MDAnalysis have to be provided, e.g. mdtraj. For this, we need infrastructure...

When accessing remote data, we need a possibility for sequential and random access for both binary and text files. For small files, transferring the whole file should be faster than seeking.

Common protocols are SFTP (paramiko) and HTTP (requests). Using rclone for further protocol support should be fine.

Steps:

1. For each protocol, identify seek and stream access.
2. Implement a wrapper class in python that behaves like a file object. Focus on read access first.
3. Implement an open call that returns the wrapper object.

When using a JSON database, iago fails to recognise the desired output format which ultimately makes any analyser script to fail.

When analysing a bucket, we can have several runs prepared with different codes. iago has to tell them apart.

Steps:

1. Analyser: make candidate file names configurable for user (for topology, input and log files)
2. Parser: Discover files, pass first 50 lines to each available reader until one of them claims the file
3. Reader: Add functionality to claim files for all available readers.