2
0
mirror of https://expo.survex.com/repositories/troggle/.git synced 2024-11-22 07:11:52 +00:00
troggle/core/utils.py

273 lines
11 KiB
Python

import logging
import random
import resource
import subprocess
from decimal import getcontext
from pathlib import Path
getcontext().prec = 2 # use 2 significant figures for decimal calculations
import settings
"""This file declares TROG a globally visible object for caches.
TROG is a dictionary holding globally visible indexes and cache functions.
It is a Global Object, see https://python-patterns.guide/python/module-globals/
troggle.utils.TROG
chaosmonkey(n) - used by survex import to regenerate some .3d files
save_carefully() - core function that saves troggle objects in the database
various git add/commit functions that need refactoring together
NOTE that TROG is not serialized! Two users can update it and conflict !!
This needs to be in a multi-user database with transactions. However it is
useful when doing a data import with databaseReset.py as that has a single
thread.
"""
TROG = {"pagecache": {"expedition": {}}, "caves": {"gcavelookup": {}, "gcavecount": {}}}
# This is module-level executable. This is a Bad Thing. Especially when it touches the file system.
try:
logging.basicConfig(level=logging.DEBUG, filename=settings.LOGFILE, filemode="w")
except:
# Opening of file for writing is going to fail currently, so decide it doesn't matter for now
pass
def get_process_memory():
usage = resource.getrusage(resource.RUSAGE_SELF)
return usage[2] / 1024.0
def chaosmonkey(n):
"""returns True once every n calls - randomly"""
if random.randrange(0, n) != 0:
return False
# print("CHAOS strikes !", file=sys.stderr)
return True
def only_commit(fname, message):
"""Only used to commit a survex file edited and saved in view/survex.py"""
git = settings.GIT
cwd = fname.parent
filename = fname.name
# print(f'{fname=} ')
try:
cp_add = subprocess.run([git, "add", filename], cwd=cwd, capture_output=True, text=True)
if cp_add.returncode != 0:
msgdata = f"Ask a nerd to fix this problem in only_commit().\n--{cp_add.stderr}\n--{cp_add.stdout}\n--return code:{str(cp_add.returncode)}"
raise WriteAndCommitError(
f"CANNOT git ADD on server for this file {filename}. Edits saved but not added to git.\n\n" + msgdata
)
cp_commit = subprocess.run([git, "commit", "-m", message], cwd=cwd, capture_output=True, text=True)
# This produces return code = 1 if it commits OK, but when the local repo still needs to be pushed to origin/loser
# which will be the case when running a test troggle system on a development machine
devok_text = """On branch master
Your branch is ahead of 'origin/master' by 1 commit.
(use "git push" to publish your local commits)
nothing to commit, working tree clean
"""
if cp_commit.returncode == 1 and cp_commit.stdout == devok_text:
pass
else:
if cp_commit.returncode != 0 and not cp_commit.stdout.strip().endswith(
"nothing to commit, working tree clean"
):
msgdata = f'--Ask a nerd to fix this problem in only_commit().\n--{cp_commit.stderr}\n--"{cp_commit.stdout}"\n--return code:{str(cp_commit.returncode)}'
print(msgdata)
raise WriteAndCommitError(
f"Error code with git on server for this file {filename}. Edits saved, added to git, but NOT committed.\n\n"
+ msgdata
)
except subprocess.SubprocessError:
raise WriteAndCommitError(
f"CANNOT git COMMIT on server for this file {filename}. Subprocess error. Edits not saved.\nAsk a nerd to fix this."
)
def write_and_commit(files, message):
"""Writes the content to the filepath and adds and commits the file to git. If this fails, a WriteAndCommitError is raised.
This does not create any needed intermediate folders, which is what we do when writing survex files, so functionality here
is duplicated in only_commit()
These need refactoring
"""
git = settings.GIT
try:
for filepath, content, encoding in files:
cwd = filepath.parent
filename = filepath.name
# GIT see also core/views/uploads.py dwgupload()
# GIT see also core/views/expo.py editexpopage()
if encoding:
mode = "w"
kwargs = {"encoding": encoding}
else:
mode = "wb"
kwargs = {}
try:
with open(filepath, mode, **kwargs) as f:
print(f"WRITING {cwd}---{filename} ")
# as the wsgi process www-data, we have group write-access but are not owner, so cannot chmod.
# os.chmod(filepath, 0o664) # set file permissions to rw-rw-r--
f.write(content)
except PermissionError:
raise WriteAndCommitError(
f"CANNOT save this file.\nPERMISSIONS incorrectly set on server for this file {filename}. Ask a nerd to fix this."
)
cp_diff = subprocess.run([git, "diff", filename], cwd=cwd, capture_output=True, text=True)
if cp_diff.returncode == 0:
cp_add = subprocess.run([git, "add", filename], cwd=cwd, capture_output=True, text=True)
if cp_add.returncode != 0:
msgdata = (
"Ask a nerd to fix this.\n\n"
+ cp_add.stderr
+ "\n\n"
+ cp_add.stdout
+ "\n\nreturn code: "
+ str(cp_add.returncode)
)
raise WriteAndCommitError(
f"CANNOT git on server for this file {filename}. Edits saved but not added to git.\n\n"
+ msgdata
)
else:
print(f"No change {filepath}")
subprocess.run([git, "commit", filename, "-m", message], cwd=cwd, capture_output=True, text=True)
cp_status = subprocess.run([git, "status", filename], cwd=cwd, capture_output=True, text=True)
#This produces return code = 1 if it commits OK, but when the repo still needs to be pushed to origin/expoweb
if cp_status.stdout.split("\n")[-2] != "nothing to commit, working tree clean":
msgdata = (
"Ask a nerd to fix this.\n\n"
+ cp_status.stderr
+ "\n\n"
+ cp_status.stdout
+ "\n\nreturn code: "
+ str(cp_status.returncode)
)
raise WriteAndCommitError(
f"Error code with git on server for this file {filename}. Edits saved, added to git, but NOT committed.\n\n"
+ msgdata
)
except subprocess.SubprocessError:
raise WriteAndCommitError(
f"CANNOT git on server for this file {filename}. Subprocess error. Edits not saved.\nAsk a nerd to fix this."
)
class WriteAndCommitError(Exception):
"""Exception class for errors writing files and comitting them to git"""
def __init__(self, message):
self.message = message
def __str__(self):
return f"WriteAndCommitError: {self.message}"
def writetrogglefile(filepath, filecontent):
"""Commit the new saved file to git
Callers to cave.writeDataFile() or entrance.writeDataFile() should handle the exception PermissionsError explicitly
"""
# GIT see also core/views/expo.py editexpopage()
# GIT see also core/views/uploads.py dwgupload()
# Called from core/models/caves.py Cave.writeDataFile() Entrance.writeDataFile()
filepath = Path(filepath)
cwd = filepath.parent
filename = filepath.name
git = settings.GIT
# as the wsgi process www-data, we have group write-access but are not owner, so cannot chmod.
# do not trap exceptions, pass them up to the view that called this function
print(f"WRITING{cwd}---{filename} ")
with open(filepath, "w") as f:
f.write(filecontent)
# os.chmod(filepath, 0o664) # set file permissions to rw-rw-r--
sp = subprocess.run([git, "add", filename], cwd=cwd, capture_output=True, check=True, text=True)
if sp.returncode != 0:
out = sp.stdout
if len(out) > 160:
out = out[:75] + "\n <Long output curtailed>\n" + out[-75:]
print(f"git ADD {cwd}:\n\n" + str(sp.stderr) + "\n\n" + out + "\n\nreturn code: " + str(sp.returncode))
sp = subprocess.run(
[git, "commit", "-m", f"Troggle online: cave or entrance edit -{filename}"],
cwd=cwd,
capture_output=True,
check=True,
text=True,
)
if sp.returncode != 0:
out = sp.stdout
if len(out) > 160:
out = out[:75] + "\n <Long output curtailed>\n" + out[-75:]
print(f"git COMMIT {cwd}:\n\n" + str(sp.stderr) + "\n\n" + out + "\n\nreturn code: " + str(sp.returncode))
# not catching and re-raising any exceptions yet, inc. the stderr etc.,. We should do that.
def save_carefully(objectType, lookupAttribs={}, nonLookupAttribs={}):
"""Looks up instance using lookupAttribs and carries out the following:
-if instance does not exist in DB: add instance to DB, return (new instance, True)
-if instance exists in DB and was modified using Troggle: do nothing, return (existing instance, False)
-if instance exists in DB and was not modified using Troggle: overwrite instance, return (instance, False)
The checking is accomplished using Django's get_or_create and the new_since_parsing boolean field
defined in core.models.TroggleModel.
We are not using new_since_parsing - it is a fossil from Aaron Curtis's design in 2006. So it is always false.
NOTE: this takes twice as long as simply creating a new object with the given values.
As of Jan.2023 this function is not used anywhere in troggle.
"""
try:
instance, created = objectType.objects.get_or_create(defaults=nonLookupAttribs, **lookupAttribs)
except:
print(" !! - FAIL in SAVE CAREFULLY ===================", objectType)
print(" !! - -- objects.get_or_create()")
print(f" !! - lookupAttribs:{lookupAttribs}\n !! - nonLookupAttribs:{nonLookupAttribs}")
raise
if not created and not instance.new_since_parsing:
for k, v in list(
nonLookupAttribs.items()
): # overwrite the existing attributes from the logbook text (except date and title)
setattr(instance, k, v)
try:
instance.save()
except:
print(" !! - SAVE CAREFULLY ===================", objectType)
print(" !! - -- instance.save()")
print(f" !! - lookupAttribs:{lookupAttribs}\n !! - nonLookupAttribs:{nonLookupAttribs}")
raise
try:
str(instance)
except:
pass
if created:
logging.info(str(instance) + " was just added to the database for the first time. \n")
if not created and instance.new_since_parsing:
logging.info(
str(instance) + " has been modified using Troggle since parsing, so the current script left it as is. \n"
)
if not created and not instance.new_since_parsing:
logging.info(
" instance:<"
+ str(instance)
+ "> existed in the database unchanged since last parse. It have been overwritten."
)
return (instance, created)