path: root/scripts/qc_on_rqtl2_bundle.py

"""Run Quality Control checks on R/qtl2 bundle."""
import sys
import json
from time import sleep
from pathlib import Path
from zipfile import ZipFile
from functools import partial
from argparse import Namespace
import multiprocessing as mproc
from logging import Logger, getLogger, StreamHandler
from typing import Union, Sequence, Callable, Iterator

import MySQLdb as mdb
from redis import Redis

from quality_control.errors import InvalidValue
from quality_control.checks import decimal_points_error

from uploader import jobs
from uploader.db_utils import database_connection
from uploader.check_connections import check_db, check_redis

from r_qtl import errors as rqe
from r_qtl import r_qtl2 as rqtl2
from r_qtl import r_qtl2_qc as rqc
from r_qtl import fileerrors as rqfe

from scripts.process_rqtl2_bundle import parse_job
from scripts.redis_logger import setup_redis_logger
from scripts.cli_parser import init_cli_parser, add_global_data_arguments

def dict2tuple(dct: dict) -> tuple:
    """Utility to convert items in dicts to pairs of tuples."""
    return tuple((key, val) for key,val in dct.items())

def add_to_errors(rconn: Redis,
                  fqjobid: str,
                  key: str,
                  errors: Sequence[Union[InvalidValue, rqfe.MissingFile]]):
    """Add `errors` to a given list of errors"""
    errs = tuple(dict(item) for item in set(
        [dict2tuple(old) for old in
         json.loads(rconn.hget(fqjobid, key) or "[]")] +
        [dict2tuple({"type": type(error).__name__, **error._asdict()})
         for error in errors]))
    rconn.hset(fqjobid, key, json.dumps(errs))

def qc_missing_files(rconn: Redis,
                     fqjobid: str,
                     zfile: ZipFile,
                     logger: Logger) -> bool:
    """Run QC for files listed in control file that don't exist in bundle."""
    logger.info("Checking for missing files…")
    missing = rqc.missing_files(zfile)
    add_to_errors(rconn, fqjobid, "errors-generic", tuple(
        rqfe.MissingFile(
            mfile[0], mfile[1], (
                f"File '{mfile[1]}' is listed in the control file under "
                f"the '{mfile[0]}' key, but it does not actually exist in "
                "the bundle."))
        for mfile in missing))
    if len(missing) > 0:
        logger.error("Missing files in the bundle!")
        return True
    return False

def compute_filesize(zfile: ZipFile, filetype: str) -> int:
    """Compute the total file size."""
    cdata = rqtl2.control_data(zfile)
    if isinstance(cdata[filetype], str):
        return zfile.getinfo(cdata[filetype]).file_size

    return sum(zfile.getinfo(afile).file_size for afile in cdata[filetype])

def retrieve_errors_with_progress(rconn: Redis,#pylint: disable=[too-many-locals]
                                  fqjobid: str,
                                  zfile: ZipFile,
                                  filetype: str,
                                  checkers: tuple[Callable]) -> Iterator[Union[
                                      InvalidValue, rqfe.MissingFile]]:
    """Filter the errors while also counting the number of lines in the file."""
    assert filetype in rqtl2.FILE_TYPES, f"Invalid file type {filetype}."
    count = 0
    checked = 0
    cdata = rqtl2.control_data(zfile)
    rconn.hset(fqjobid, f"{filetype}-filesize", compute_filesize(zfile, filetype))
    def __update_processed__(value):
        nonlocal checked
        checked = checked + len(value)
        rconn.hset(fqjobid, f"{filetype}-checked", checked)

    try:# pylint: disable=[too-many-nested-blocks]
        for lineno, row in enumerate(
                rqtl2.file_data(zfile, filetype, cdata), start=1):
            count = count + 1
            for field, value in row.items():
                if field == "id":
                    __update_processed__(value)
                    continue
                if value is not None:
                    for checker in checkers:
                        error = checker(lineno=lineno, field=field, value=value)
                        if bool(error):
                            yield error
                        __update_processed__(value)

        rconn.hset(fqjobid, f"{filetype}-linecount", count)
    except rqe.MissingFileError:
        fname = cdata.get(filetype)
        yield rqfe.MissingFile(filetype, fname, (
            f"The file '{fname}' does not exist in the bundle despite it being "
            f"listed under '{filetype}' in the control file."))

def qc_geno_errors(rconn, fqjobid, _dburi, _speciesid, zfile, logger) -> bool:
    """Check for errors in `geno` file(s)."""
    cdata = rqtl2.control_data(zfile)
    if "geno" in cdata:
        logger.info("Checking for errors in the 'geno' file…")
        gerrs = tuple(retrieve_errors_with_progress(
            rconn, fqjobid, zfile, "geno",
            (rqc.make_genocode_checker(cdata.get("genotypes", {}), "geno"),)))
        add_to_errors(rconn, fqjobid, "errors-generic", tuple(
            err for err in gerrs if isinstance(err, rqfe.MissingFile)))
        add_to_errors(rconn, fqjobid, "errors-geno", tuple(
            err for err in gerrs if not isinstance(err, rqfe.MissingFile)))
        if len(gerrs) > 0:
            logger.error("The 'geno' file has errors.")
            return True
        logger.info("No errors found in the 'geno' file.")

    return False


def fetch_db_geno_samples(conn: mdb.Connection, speciesid: int) -> tuple[str, ...]:
    """Fetch samples/cases/individuals from the database."""
    samples = set()# type: ignore[var-annotated]
    with conn.cursor() as cursor:
        cursor.execute("SELECT Name, Name2 from Strain WHERE SpeciesId=%s",
                       (speciesid,))
        rows = cursor.fetchall() or tuple()
        for row in rows:
            samples.update(tuple(row))

    return tuple(item.strip() for item in samples if bool(item))


def check_pheno_samples(
        conn: mdb.Connection,
        speciesid: int,
        zipfilepath: Union[str, Path],
        logger: Logger
) -> tuple[Union[InvalidValue, rqfe.MissingFile], ...]:
    """Check that samples in 'pheno' file exist in geno file."""
    cdata = rqtl2.read_control_file(zipfilepath)
    genosamples = tuple(
        sample for perfilesamples in (
            rqtl2.load_samples(zipfilepath, member, cdata["geno_transposed"])
            for member in cdata["geno"])
        for sample in perfilesamples) + fetch_db_geno_samples(conn, speciesid)

    def __check_file__(member) -> tuple[InvalidValue, ...]:
        logger.info("Checking samples/cases in member file '%s' …", member)
        sampledata = rqtl2.load_samples(
            zipfilepath, member, cdata["pheno_transposed"])
        errors: tuple[InvalidValue, ...] = tuple()
        for sample in sampledata:
            if sample not in genosamples:
                errors = errors + (InvalidValue(
                    member, "-", "-", sample,
                    f"The individual/case/sample '{sample}' in file "
                    f"{member} does not exist in either, any of the 'geno' "
                    "files provided in the bundle or the GeneNetwork database."
                ),)

        logger.info("Found %s missing samples in member file '%s'.",
                    len(errors),
                    member)
        return errors

    allerrors: tuple[Union[InvalidValue, rqfe.MissingFile], ...] = tuple()
    for afile in cdata["pheno"]:
        try:
            allerrors = allerrors + __check_file__(afile)
        except KeyError:
            allerrors = allerrors + (rqfe.MissingFile(
                "pheno",
                afile,
                (f"The file '{afile}' does not exist in the zipfile despite "
                 "being listed in the control file.")),)

    return allerrors


def qc_pheno_errors(# pylint: disable=[too-many-arguments]
        rconn, fqjobid, dburi, speciesid, zfile, logger) -> bool:
    """Check for errors in `pheno` file(s)."""
    cdata = rqtl2.control_data(zfile)
    if "pheno" in cdata:
        logger.info("Checking for errors in the 'pheno' file…")
        perrs = tuple()# type: ignore[var-annotated]
        with database_connection(dburi) as dbconn:
            perrs = check_pheno_samples(
                dbconn, speciesid, zfile.filename, logger) + tuple(
                    retrieve_errors_with_progress(
                        rconn,fqjobid, zfile, "pheno",
                        (partial(decimal_points_error,
                                 filename="pheno",
                                 mini=3),)))
        add_to_errors(rconn, fqjobid, "errors-generic", tuple(
            err for err in perrs if isinstance(err, rqfe.MissingFile)))
        add_to_errors(rconn, fqjobid, "errors-pheno", tuple(
            err for err in perrs if not isinstance(err, rqfe.MissingFile)))
        if len(perrs) > 0:
            logger.error("The 'pheno' file has errors.")
            return True
        logger.info("No errors found in the 'pheno' file.")

    return False

def qc_phenose_errors(# pylint: disable=[too-many-arguments]
        rconn, fqjobid, _dburi, _speciesid, zfile, logger) -> bool:
    """Check for errors in `phenose` file(s)."""
    cdata = rqtl2.control_data(zfile)
    if "phenose" in cdata:
        logger.info("Checking for errors in the 'phenose' file…")
        perrs = tuple(retrieve_errors_with_progress(
            rconn,fqjobid, zfile, "phenose",
            (partial(decimal_points_error, mini=6),)))
        add_to_errors(rconn, fqjobid, "errors-generic", tuple(
            err for err in perrs if isinstance(err, rqfe.MissingFile)))
        add_to_errors(rconn, fqjobid, "errors-phenose", tuple(
            err for err in perrs if not isinstance(err, rqfe.MissingFile)))
        if len(perrs) > 0:
            logger.error("The 'phenose' file has errors.")
            return True
        logger.info("No errors found in the 'phenose' file.")

    return False

def qc_phenocovar_errors(
        _rconn,
        _fqjobid,
        _dburi,
        _speciesid,
        _zfile,
        _logger
) -> bool:
    """Check for errors in `phenocovar` file(s)."""
    return False

def run_qc(rconn: Redis,
           args: Namespace,
           logger: Logger) -> int:
    """Run the QC programs."""
    fqjobid = jobs.job_key(args.redisprefix, args.jobid)
    thejob = parse_job(rconn, args.redisprefix, args.jobid)
    jobmeta = thejob["job-metadata"]

    with ZipFile(jobmeta["rqtl2-bundle-file"], "r") as zfile:
        if qc_missing_files(rconn, fqjobid, zfile, logger):
            return 1

    def with_zipfile(# pylint: disable=[too-many-arguments]
            rconn, fqjobid, dbconn, speciesid, filename, logger, func
    ):
        with ZipFile(filename, "r") as zfile:
            return func(rconn, fqjobid, dbconn, speciesid, zfile, logger)

    def buildargs(func):
        return (rconn,
                fqjobid,
                args.databaseuri,
                args.speciesid,
                jobmeta["rqtl2-bundle-file"],
                logger,
                func)
    processes = [
        mproc.Process(target=with_zipfile, args=buildargs(qc_geno_errors,)),
        mproc.Process(target=with_zipfile, args=buildargs(qc_pheno_errors,)),
        mproc.Process(target=with_zipfile, args=buildargs(qc_phenose_errors,)),
        mproc.Process(target=with_zipfile, args=buildargs(qc_phenocovar_errors,))
    ]
    for process in processes:
        process.start()

    while True:
        processes_running = any(
            (process.is_alive() for process in processes))
        if not processes_running:
            break
        sleep(2)

    if any((process.exitcode for process in processes)):
        # at least one process failed for some reason...
        return 1

    def __fetch_errors__(rkey: str) -> tuple:
        return tuple(json.loads(rconn.hget(fqjobid, rkey) or "[]"))

    return (1 if any((
        bool(__fetch_errors__(key))
        for key in
        ("errors-geno", "errors-pheno", "errors-phenos", "errors-phenocovar")))
            else 0)

if __name__ == "__main__":
    def main():
        """Enter R/qtl2 bundle QC runner."""
        args = add_global_data_arguments(init_cli_parser(
            "qc-on-rqtl2-bundle", "Run QC on R/qtl2 bundle.")).parse_args()
        check_redis(args.redisuri)
        check_db(args.databaseuri)

        logger = getLogger("qc-on-rqtl2-bundle")
        logger.addHandler(StreamHandler(stream=sys.stderr))
        logger.setLevel("DEBUG")

        fqjobid = jobs.job_key(args.redisprefix, args.jobid)
        with Redis.from_url(args.redisuri, decode_responses=True) as rconn:
            logger.addHandler(setup_redis_logger(
                rconn, fqjobid, f"{fqjobid}:log-messages",
                args.redisexpiry))

            exitcode = run_qc(rconn, args, logger)
            rconn.hset(
                jobs.job_key(args.redisprefix, args.jobid), "exitcode", exitcode)
            return exitcode

    sys.exit(main())