# Feature: logging support
**URL:** https://discourse.vtk.org/t/feature-logging-support/171
**Category:** Development
**Created:** [January 30, 2019, 3:35pm UTC](https://discourse.vtk.org/t/feature-logging-support/171 "2019-01-30T15:35:50Z")
**Posts on this page:** 20
**Page:** 1
### Author:  [@utkarshayachit](https://discourse.vtk.org/u/utkarshayachit)
#### Post date: [January 30, 2019, 3:35pm UTC](https://discourse.vtk.org/t/feature-logging-support/171/1 "2019-01-30T15:35:50Z")
This is a proposal with draft implementation to add support for a logging framework to VTK originally introduce in [this mailing list thread](https://vtk.org/pipermail/vtk-developers/2018-August/036256.html).
Here’s an extract from the header docs in [vtkLogger.h](https://gitlab.kitware.com/utkarsh.ayachit/vtk/tree/advanced-logging/Common/Core/vtkLogger.h) that summarizes how to use this logging capability.
```auto
/// Optional, but useful to time-stamp the start of the log.
// Will also detect verbosity level on the command line as -v.
vtkLogger::Init(argc, argv);
// Put every log message in "everything.log":
vtkLogger::LogToFile("everything.log", vtkLogger::APPEND, vtkLogger::VERBOSITY_MAX);
// Only log INFO, WARNING, ERROR to "latest_readable.log":
vtkLogger::LogToFile("latest_readable.log", vtkLogger::TRUNCATE, vtkLogger::VERBOSITY_INFO);
// Only show most relevant things on stderr:
vtkLogger::SetStderrVerbosity(vtkLogger::VERBOSITY_1);
// add a line to log using the verbosity name.
vtkLogF(INFO, "I'm hungry for some %.3f!", 3.14159);
vtkLogF(0, "same deal");
// add a line to log using the verbosity enum.
vtkVLogF(vtkLogger::VERBOSITY_INFO, "I'm hungry for some %.3f!", 3.14159);
vtkVLogF(vtkLogger::VERBOSITY_0, "same deal");
// to add a indentifier for a vtkObjectBase or subclass
vtkLogF(INFO, "The object is %s", vtkLogIdentifier(vtkobject));
// add a line conditionally to log if the condition is true:
vtkLogIfF(INFO, ptr == nullptr, "ptr is nullptr (some number: %.3f)", * 3.14159);
vtkLogScopeF(INFO, "Will indent all log messages within this scope.");
// in a function, you may use vtkLogScopeFunction(INFO)
// alternatively, you can use streams instead of printf-style
vtkLog(INFO, "I'm hungry for some " << 3.14159 << "!");
vtkLogIf(INFO, ptr != nullptr, "ptr is " << "nullptr");
```
A few things to note:
- based on loguru 2.0.0 however we expose VTK specific APIs and macros to support future replacing underlying implementation, if needed
- supports printf-style and streams via `vtkLog..()` are stream based macros, while `vtkLog..F()` are their printf-style counter parts.
- CMake flag available to disable logging entirely at which point all these macros simply become no-ops.
- level based logging with support for multiple outputs with different verbosity levels.
- `vtkErrorMacro`, `vtkWarningMacro` etc. automatically log the messages as well (without the extra filename, line number info since that is already captured by the log more cleanly). It’s possible to bypass the vtkOutputWindow based legacy support entirely and just rely on logging by calling `vtkObject::GlobalWarningDisplayOff()`. Additionally, to avoid duplicate messages on the terminal, when VTK is built with logging support vtkOutputWindow doesn’t output the messages to the `stdout` or `stderr`. It will still fire the appropriate message signals, just not show them on the standard output streams to avoid messing up with the log stream.
- loguru supports for adding handlers/callbacks; currently working on adding VTK exposed API for the same. The plan is make it available for developers so developers can capture messages being logged for doing things like reporting in an message window etc.
- loguru supports ability to format log generated; currently working on VTK-exposed API for the same which will be finalized in time for the merge (or soon after)
The changes are being developed/tested here: [https://gitlab.kitware.com/vtk/vtk/merge\_requests/5132](https://gitlab.kitware.com/vtk/vtk/merge_requests/5132)
Any feedback, suggestions, comments are welcome as always.
Edits:
- changed class name to `vtkLogger` from `vtkLog`
- changed macro case / style as suggested.
- changed to use `F` suffix instead of `f` for format overloads
- corrected docs for `vtkLogIfF`
---
### Author:  [@cory.quammen](https://discourse.vtk.org/u/cory.quammen)
#### Post date: [January 30, 2019, 10:24pm UTC](https://discourse.vtk.org/t/feature-logging-support/171/2 "2019-01-30T22:24:26Z")
In keeping with how most classes in VTK are named with nouns (e.g., `vtkPicker`, `vtkMapper`, `vtkXMLReader`), I wonder if the class should be named `vtkLogger` instead of `vtkLogging`?
---
### Author:  [@utkarshayachit](https://discourse.vtk.org/u/utkarshayachit)
#### Post date: [January 31, 2019, 3:01pm UTC](https://discourse.vtk.org/t/feature-logging-support/171/3 "2019-01-31T15:01:32Z")
> [@cory.quammen](#):
>
> In keeping with how most classes in VTK are named with nouns (e.g., `vtkPicker` , `vtkMapper` , `vtkXMLReader` ), I wonder if the class should be named `vtkLogger` instead of `vtkLogging` ?
fair point. I’ll change it shortly.
---
### Author:  [@cryos](https://discourse.vtk.org/u/cryos)
#### Post date: [January 31, 2019, 5:06pm UTC](https://discourse.vtk.org/t/feature-logging-support/171/4 "2019-01-31T17:06:59Z")
Why all caps and underscores for LOGGING\_CALLS? The old methods followed standard naming conventions/capitalization, visually it looks like a step backwards. I guess they are macros, hence the \_F and \_S, could we consider vtkLogString and vtkLogStream to be a little more explicit and less screamy?
---
### Author:  [@utkarshayachit](https://discourse.vtk.org/u/utkarshayachit)
#### Post date: [January 31, 2019, 5:17pm UTC](https://discourse.vtk.org/t/feature-logging-support/171/5 "2019-01-31T17:17:49Z")
> [@cryos](#):
>
> Why all caps and underscores for LOGGING\_CALLS
you’re correct, because they are macros and that was the style loguru & VTK-m used.
> [@cryos](#):
>
> could we consider vtkLogString and vtkLogStream to be a little more explicit and less screamy?
Sure. Given VTK tradition, how about we use `vtkLog(...)` for stream-based variants (as I suspect that’d be more common) and use `vtkLogf(...)` for the `printf` variants?
---
### Author:  [@mwestphal](https://discourse.vtk.org/u/mwestphal)
#### Post date: [February 6, 2019, 8:49am UTC](https://discourse.vtk.org/t/feature-logging-support/171/6 "2019-02-06T08:49:57Z")
Nice!
How does one enable/configure vtkLogging in their own vtk application?
---
### Author:  [@utkarshayachit](https://discourse.vtk.org/u/utkarshayachit)
#### Post date: [February 6, 2019, 1:34pm UTC](https://discourse.vtk.org/t/feature-logging-support/171/7 "2019-02-06T13:34:35Z")
> [@mwestphal](#):
>
> How does one enable/configure vtkLogging in their own vtk application?
See docs in [vtkLogger.h](https://gitlab.kitware.com/vtk/vtk/blob/master/Common/Core/vtkLogger.h#L66-101), or [TestLogger.cxx](https://gitlab.kitware.com/vtk/vtk/blob/master/Common/Core/Testing/Cxx/TestLogger.cxx).
---
### Author:  [@allison.vacanti](https://discourse.vtk.org/u/allison.vacanti)
#### Post date: [February 11, 2019, 2:49pm UTC](https://discourse.vtk.org/t/feature-logging-support/171/8 "2019-02-11T14:49:42Z")
I’m a little late to the discussion, but since vtk-m is already using loguru, would it be possible to share the same backend and log from both to the same file?
[https://gitlab.kitware.com/vtk/vtk-m/blob/master/vtkm/cont/Logging.h](https://gitlab.kitware.com/vtk/vtk-m/blob/master/vtkm/cont/Logging.h)
---
### Author:  [@utkarshayachit](https://discourse.vtk.org/u/utkarshayachit)
#### Post date: [February 11, 2019, 2:55pm UTC](https://discourse.vtk.org/t/feature-logging-support/171/9 "2019-02-11T14:55:16Z")
I am afraid not. VTK-m is not a required dependency of VTK or vice-versa. In the long run, when VTK-m becomes a required dependency we can revisit. Right now, both loguru’s are mangled separately and hence will end up in different logs – is my guess.
---
### Author:  [@estan](https://discourse.vtk.org/u/estan)
#### Post date: [February 11, 2019, 5:46pm UTC](https://discourse.vtk.org/t/feature-logging-support/171/10 "2019-02-11T17:46:29Z")
This looks nice, but from the example:
> [@utkarshayachit](#):
>
> ```auto
> // add a line conditionally to log if the condition fails:
> vtkLogIff(INFO, ptr != nullptr, "ptr is nullptr (some number: %.3f)", * 3.14159);
>
> ```
I found this quite confusing when I read it. The function name is `vtkLogIff` and takes a condition, but it will log if the condition _doesn’t_ hold? If so, I would call it `vtkLogIfNot` or `vtkLogUnless`, and maybe have a `vtkLogIf` that does the opposite for those who prefer that style.
The way it is, I read `vtkLogIff(INFO, ptr != nullptr, ...)` as “log if ptr is not nullptr” which is opposite to what the function seems to do.
Also, the `vtkLogIff` with the extra `f` (I guess for “format”) is a bit unfortunate, since I think many might read that as “if and only if”, adding even more to the confusion? EDIT: One way to avoid it is to name it `vtkLogIfF`. I know it’s sort of ugly, but it does makes sense if one follows camel case and consider that the `F` is short for `Formatted`.
Or perhaps I’m just a confused person 🙂
---
### Author:  [@utkarshayachit](https://discourse.vtk.org/u/utkarshayachit)
#### Post date: [February 11, 2019, 5:56pm UTC](https://discourse.vtk.org/t/feature-logging-support/171/11 "2019-02-11T17:56:51Z")
> [@estan](#):
>
> The function name is `vtkLogIff` and takes a condition, but it will log if the condition _doesn’t_ hold?
good point. I think I will change it to _log if condition is true_, which seems a more natural way – incidentally, that’s how it’s implemented, I just had it wrong in the docs – I’ll update it.
The `vtkLogIff` is indeed a unfortunate occurrence. The lower `f` was chosen to match `printf`, but by strict VTK style, `-f` should indeed be capitalized. Unless there’s objection I will change that accordingly as well.
fixes included [here](https://gitlab.kitware.com/vtk/vtk/merge_requests/5176)
---
### Author:  [@jal](https://discourse.vtk.org/u/jal)
#### Post date: [February 14, 2019, 8:17pm UTC](https://discourse.vtk.org/t/feature-logging-support/171/12 "2019-02-14T20:17:00Z")
IDK if this is related, but i have been getting runtime errors from loguru when running examples today:
To run a test, enter the test number: 54
( 3.611s) [] loguru.cpp:529 FATL| CHECK FAILED: argc \> 0 (0 \> 0) Expected proper argc/argv
I see this on windows and linux in the master branch. Any ideas what is wrong?
Best
Jeff
---
### Author:  [@utkarshayachit](https://discourse.vtk.org/u/utkarshayachit)
#### Post date: [February 14, 2019, 8:23pm UTC](https://discourse.vtk.org/t/feature-logging-support/171/13 "2019-02-14T20:23:10Z")
looks like the examples are passing malformed argc,argv. `loguru::init(..)` confirms that `argv[argc] == nullptr` and `argc > 0`. what’s the callstack?
---
### Author:  [@jal](https://discourse.vtk.org/u/jal)
#### Post date: [February 14, 2019, 8:31pm UTC](https://discourse.vtk.org/t/feature-logging-support/171/14 "2019-02-14T20:31:12Z")
i remember it was in vtkLogger::Init, ill have to wait until finished build to get exact.
---
### Author:  [@jal](https://discourse.vtk.org/u/jal)
#### Post date: [February 14, 2019, 8:35pm UTC](https://discourse.vtk.org/t/feature-logging-support/171/15 "2019-02-14T20:35:56Z")
i got one on linux…
Stack trace:
4 0x4421d9 vtkCommonCoreCxxTests() [0x4421d9]
3 0x7fb83746ed20 \_\_libc\_start\_main + 256
2 0x442668 vtkCommonCoreCxxTests() [0x442668]
1 0x7fb836197275 vtkLogger::Init(int&, char\*\*, char const\*) + 63
0 0x7fb83485777c vtkloguru::init(int&, char\*\*, char const\*) + 295
( 1.668s) [379ED720] loguru.cpp:529 FATL| CHECK FAILED: argc \> 0 (0 \> 0) Expected proper argc/argv
---
### Author:  [@utkarshayachit](https://discourse.vtk.org/u/utkarshayachit)
#### Post date: [February 14, 2019, 8:37pm UTC](https://discourse.vtk.org/t/feature-logging-support/171/16 "2019-02-14T20:37:30Z")
totally strange, what’s the name of the test that this is failing for?
---
### Author:  [@jal](https://discourse.vtk.org/u/jal)
#### Post date: [February 14, 2019, 9:25pm UTC](https://discourse.vtk.org/t/feature-logging-support/171/17 "2019-02-14T21:25:18Z")
seems to be all tests on both linux and windows. i’m trying a clean clone…
---
### Author:  [@allison.vacanti](https://discourse.vtk.org/u/allison.vacanti)
#### Post date: [February 18, 2019, 10:43am UTC](https://discourse.vtk.org/t/feature-logging-support/171/18 "2019-02-18T10:43:26Z")
> CHECK FAILED: argc \> 0 (0 \> 0) Expected proper argc/argv
Is there code calling `loguru::init(0, nullptr);` somewhere? We ran into this on vtk-m, and it takes a bit of work to fool loguru into taking a dummy argv/argc:
```auto
int argc = 1;
char dummy[1] = { '\0' };
char* argv[2] = { dummy, nullptr };
loguru::init(argc, argv);
```
[https://gitlab.kitware.com/vtk/vtk-m/blob/master/vtkm/cont/Logging.cxx#L148-154](https://gitlab.kitware.com/vtk/vtk-m/blob/master/vtkm/cont/Logging.cxx#L148-154)
---
### Author:  [@jal](https://discourse.vtk.org/u/jal)
#### Post date: [February 18, 2019, 6:53pm UTC](https://discourse.vtk.org/t/feature-logging-support/171/19 "2019-02-18T18:53:55Z")
Thanks Utkarsh, Allison,
sorry for late reply, i am still building and will let you know if the problem is still there… if you are not seeing it when running the examples like vtkCommonCoreCxxTests, then it is probably an issue with my build.
---
### Author:  [@jal](https://discourse.vtk.org/u/jal)
#### Post date: [February 19, 2019, 1:34am UTC](https://discourse.vtk.org/t/feature-logging-support/171/20 "2019-02-19T01:34:39Z")
with a fresh pull and build, still the following from vtkRenderingAnnotationCxxTests: are you not seeing this?
Stack trace:
4 0x4052f9 vtkRenderingAnnotationCxxTests() [0x4052f9]
3 0x7f1d79322d20 \_\_libc\_start\_main + 256
2 0x405bc5 vtkRenderingAnnotationCxxTests() [0x405bc5]
1 0x7f1d71f72275 vtkLogger::Init(int&, char\*\*, char const\*) + 63
0 0x7f1d6b64477c vtkloguru::init(int&, char\*\*, char const\*) + 295
( 3.187s) [79761820] loguru.cpp:529 FATL| CHECK FAILED: argc \> 0 (0 \> 0) Expected proper argc/argv
[Next page](https://discourse.vtk.org/t/feature-logging-support/171.md?page=2)