Added a vignette with advice for Licor LI-6800 user constants.
Added a new function for calculating values of Jmax (calculate_jmax) and included it in the C3 and C4 A-Ci curve vignettes. Along with this function, new temperature response parameters were added: jmax_temperature_param_bernacchi and jmax_temperature_param_flat. These employ polynomial temperature responses, a new type that was also added to the package via the calculate_temperature_response_polynomial function.
identify_c3_limiting_processes now returns each co-limiting process when assimilation is co-limited.
Several functions were modified to accept data.frame objects in addition to exdf objects: calculate_ball_berry_index, calculate_c3_assimilation, calculate_residuals, document_variables, identify_c3_limiting_processes, and set_variable.
Added functionality for reading user remarks to read_licor_6800_plaintext and read_licor_6800_Excel.
Added a new input argument to organize_response_curve_data that enables calculations of average values for specified columns.
Improved exdf construction and element access; now it is possible to provide just a few units when creating an exdf object, and it is possible to create single-column exdf objects.
When calling print or str on an exdf object, the text printed to the terminal now explains that it was generated by converting an exdf to a data.frame.
Licor LI-6800 Excel log files must now have a sheet called Measurements. A Remarks sheet is optional.
factorize_id_column can now work with other control group names besides WT.
Bug fixes
A bug was fixed where Gamma_star was sometimes accidentally removed from the outputs of fit_c3_aci and fit_c3_variable_j
Optional input arguments to calculate_c3_assimilation are checked to make sure they are supported. Going forward, any new functions with optional arguments (such as calculate_jmax) will use this same check.
Removed Qin from several functions related to C4 A-Ci curves, since Qin is not actually needed for fitting C4 A-Ci curves.
Internal changes
The unit_dictionary is now a function rather than a list, which enables better error messages when a quantity is not included in the dictionary; a few mistakes were discovered this way (and fixed).
Changes in PhotoGEA version 1.1.0 (2024-12-09)
The parameter trust values have been changed from logical values (TRUE or FALSE) to a numeric indicator (0, 1, or 2). These values indicate severe unreliability, potential unreliability, and full reliability, respectively.
Added several optional input arguments to calculate_c3_assimilation that allow variants of the FvCB model to be used: use_min_A, TPU_threshold, use_FRL, and consider_depletion. These can also be specified when fitting curves.
The behavior of identify_c3_unreliable_pts was modified to make it compatible with the new use_min_A option.
A new option was added to remove_points, allowing users to exclude points from any subsequent fits rather than completely removing them.
plot_c3_aci_fit and the other fit plotting functions have been altered to distinguish between points included in or excluded from the fit.
xyplot_avg_rc has been altered to exclude such points when calculating average values
Outliers can now be excluded automatically when creating a barchart using barchart_with_errorbars
More information is now included in the output from fit_ball_berry: standard error for each fitted parameter, and the p-value for the fit. A dedicated plotting function was also added (plot_ball_berry_fit).
A function was added for fitting the Medlyn conductance model (as an alternative to the Ball-Berry model): fit_medlyn
Update input argument checks for several functions: check_required_columns and set_variable
The system for specifying temperature response parameters and calculating temperature-dependent values of key photosynthetic parameters has been been revamped to make it more flexible and easier to use:
calculate_arrhenius and calculate_peaked_gaussian have been renamed to calculate_temperature_response_arrhenius and calculate_temperature_response_gaussian to better reflect their purpose.
A new type of temperature response has been added: calculate_temperature_response_johnson.
When calculating C3 assimilation rates or fitting C3 A-Ci curves, Tp is now specified at 25 degrees C and follows a temperature response function, as was already done for Vcmax, J, and RL.
When calculating C4 assimilation rates or fitting C4 A-Ci curves, Jmax is now specified at 25 degrees C, rather than at its optimum value, for consistency with the other parameters that can be fit in PhotoGEA.
A central function for calculating temperature responses has been added: calculate_temperature_response. It internally calls the other functions, so users can just use this single function.
The c4_arrhenius_von_caemmerer and c4_peaked_gaussian_von_caemmerer have been consolidated into a single list suitable for use with calculate_temperature_response, caled c4_temperature_param_vc.
The c3_arrhenius_sharkey and c3_arrhenius_bernacchi lists now include Johnson-Eyring-Williams reponses for Tp and gmc and have been renamed to c3_temperature_param_sharkey and c3_temperature_param_bernacchi since they are no longer pure list of Arrhenius parameters.
These changes may not be compatible with scripts written for earlier versions of PhotoGEA.
read_licor_6800_Excel now checks to make sure the A and gsw columns are not all zero; if these columns are all zero, this indicates that the Excel file needs to be “calculated,” and a relevant message is sent to the user.
The C3 assimilation model and C3 A-Ci fits have been updated to include:
Another parameter related to TPU (alpha_t)
Mesophyll conductance (gmc_at_25)
The C4 A-Ci fits have also been updated to include mesophyll conductance (gmc_at_25) and to fit J rather than Jmax.
Revamped the Laisk fitting function:
Renamed the function from calculate_RL_laisk to fit_laisk since a fitting procedure is used, and the method also estimates Ci_star in addition to RL.
Used the method described in Walker & Ort (2015) and Busch et al. (2024)
Added a plotting function for displaying the results of a Laisk fit: plot_laisk_fit.
AIC calculations were moved to residual_stats so the AIC can be more easily calculated for fits from other packages.
Added a new optimizer (optimizer_nlminb) and improved the optimizer documentation and testing. The new optimizer is a wrapper for stats::nlminb and employs a quasi-Newton algorithm.
Added a new function for identifying limiting processes in C3 curves: identify_c3_limiting_processes. This function is used internally by fit_c3_aci and fit_c3_variable_j.
A new option was added to Variable J fits: check_j
Changes in PhotoGEA version 1.0.0 (2024-08-13)
A new fitting function was added: fit_c4_aci_hyperbola. This allows users to fit an empirical hyperbola to C4 A-Ci curves, rather than the mechanistic model used in fit_c4_aci.
Several supporting functions were also added: calculate_c4_assimilation_hyperbola, confidence_intervals_c4_aci_hyperbola, error_function_c4_aci_hyperbola, initial_guess_c4_aci_hyperbola, and plot_c4_aci_hyperbola_fit.
Tests were also added for fit_c4_aci_hyperbola and calculate_c4_assimilation_hyperbola
Added a new control option so users can set or bypass hard constraints on parameter values when calculating assimilation rates
The input is called hard_constraints and it takes a numerical value, where higher values impose more constraints on parameter values. The highest value is 2.
Setting hard_constraints to 2 is equivalent to the default behavior in previous versions of PhotoGEA.
The default value in all functions that take it as an input is 0, which imposes no hard constraints.
It has been added to calculate_c3_assimilation, calculate_c3_variable_j, calculate_c4_assimilation, and calculate_c4_assimilation_hyperbola, as well as any functions that use these internally, such as calculate_c3_limitations_warren and fit_c3_aci.
The default bounds for all the curve fitting functions have been expanded to avoid biasing the results.
Confidence limits for parameters at leaf temperature have been added.
Options for identifying and removing unreliable parameter estimates have been added. With this change, the remove_unreliable_param input argument must now be a numeric value rather than a logical value.
A value of 0 disables this feature (equivalent to FALSE in previous versions of PhotoGEA).
A value of 1 removes parameters (and their corresponding rates) if the corresponding rate is never the smallest rate.
A value of 2 removes parameters (and their corresponding rates) if the corresponding rate is never the smallest rate, and removes parameters if the the upper confidence limit is infinity (equivalent to TRUE in previous versions of PhotoGEA).
The default value for all functions that have this option is 2.
A warning was removed from fit_c4_aci, which had previously suggested to avoid fitting more than one of Vcmax_at_25, Vpr, and Jmax_at_opt.
The default value of the require_positive_gmc input argument of the fit_c3_variable_j function has been changed to positive_a.
The names of respiration rates were changed: Rd (the rate of day respiration) has been changed to RL (the rate of respiration in the light) and Rm (the rate of day respiration occurring in the mesophyll) has been changed to RLm (the rate of respiration in the light occurring in the mesophyll). The subscript L for “light” is more clear than d for “day,” since in some contexts d refers to “dark.” One function name was modified during this process: calculate_rd_laisk became calculate_RL_laisk.
The basic_stats function was updated so it can ignore NA values when calculating averages and standard errors.
The check_response_curve_data function was updated so there is now an option to create a warning rather than an error message when a problem is detected.
The estimate_operating_point function was updated so a value of NA for the Ca_atmospheric input simply bypasses the operating point calculations. Along with this, the default value for Ca_atmospheric was set to NA in the fit_c3_aci, fit_c3_variable_j, and fit_c4_aci functions. With these changes, calculating the operating point is now optional.
The C3 and C4 A-Ci vignettes were updated to remove some outdated information and to take advantage of the new fitting function fit_c4_aci_hyperbola.
Increased minimum required R version from 3.6.0 to 4.0.0. The GitHub actions testing setup no longer works for R < 4.0, so it has become too difficult to guarantee compatability with an earlier version.
Changes in PhotoGEA version 0.12.0 (2024-06-03)
Changed fitting method from least-squares to maximum likelihood in fit_c3_aci, fit_c3_variable_j, and fit_c4_aci.
We use a normal distribution for calculating the likelihood.
Best-fit parameter values are determined with sigma = 1.
Then the true value of the likelihood can be estimated using sigma = RMSE.
Confidence intervals are also calculated using sigma = RMSE.
Fitting functions include the Akaike information criterion (AIC) in their outputs
Default settings have been changed to always calculate confidence intervals and remove unreliable parameter estimates, to use a more reliable (but slower) optimizer, and to fit alpha_old (for C3 A-Ci and Variable J fits); these settings will ensure more robust fitting results
Changed the C3 assimilation and Variable J equations to accommodate the new alpha parameters from Busch et al. (2018)
There are now three separate parameters: alpha_old (previously alpha_g), alpha_g, and alpha_s
If alpha_old is nonzero, then alpha_g and alpha_s must be zero. Likewise, if alpha_g or alpha_s is nonzero, then alpha_old must be zero. This will prevent users from mixing the two models together.
The fitting functions fit_c3_aci, fit_c3_variable_j, and fit_c4_aci now include a new output called fits_interpolated that contains values of the predicted assimilation rates with a Ci step of 1 ppm.
New plotting functions have been added: plot_c3_aci_fit and plot_c4_aci_fit. These functions use the new information in fits_interpolated to make nice plots comparing the measured data and the fits.
Now users can optionally ignore NA values when using xyplot_avg_rc and barchart_with_errorbars
Changed exclude_outliers to make sure it doesn’t exclude NA values
Specialized functions for writing exdf objects to CSV files and recreating exdf objects from those files are now available: write.csv.exdf and read.csv.exdf.
When determining the degree of trust in a best-fit parameter value, we now consider parameters with an upper confidence limit of Inf to be unreliable.
Light- and electron-limited assimilation has been added to calculate_c4_assimilation; now we have fully implemented the von Caemmerer model equations. This also necessitated a new function for temperature response calculations: calculate_peaked_gaussian.
New fitting parameters have been added to fit_c4_aci: alpha_psii, gbs, Jmax_at_opt and Rm_frac.
It is now possible to remove unreliable parameter estimates when using fit_c4_aci.
Tests have been updated to make sure the fitting functions can gracefully handle a fit failure, even when estimating confidence intervals and/or removing unreliable parameter estimates.
Added a new optimizer (optimizer_hjkb) and changed their default arguments so a user must always specify the tolerance or number of generations.
All functions that require an O2 partial pressure now calculate it from the total pressure and the oxygen concentration (expressed as a percentage).
A new option has been added to read_licor_6800_Excel and read_licor_6800_plaintext: get_oxygen. When this input is TRUE, get_oxygen_from_preamble will automatically be used to get the oxygen percentage from the file’s preamble when it is loaded.
A new option has been added to read_gasex_file: standardize_columns.
read_licor_6800_plaintext can now read log files that were closed and reopened
Added a new function called PhotoGEA_example_file_path to avoid using system.file in examples since system.path has been confusing for some users
check_licor_data has been renamed to check_response_curve_data since it is not limited to only Licor measurements
Specified a minimum supported R version: 3.6.0.
Changes in PhotoGEA version 0.11.0 (2024-02-12)
Added new options for adding penalties to the error function during Variable J fits, that enable the user to selectively penalize negative or unreasonably large values of mesophyll conductance.
Made a few improvements to C3 curve fitting functions (fit_c3_aci and fit_c3_variable_j):
Two more parameters can now be fit: alpha (related to TPU) and Gamma_star.
Unreliable parameter estimates can now be excluded; for example, if no points on a curve have An = Ap, the fit will return NA for TPU.
The initial guess functions (initial_guess_c3_aci and initial_guess_c3_variable_j) can now accommodate user-supplied values of alpha.
Made a few improvements to all three nonlinear fitting functions (fit_c3_aci, fit_c3_variable_j, and fit_c4_aci):
The error functions are now available in the package namespace as error_function_c3_aci, error_function_c3_variable_j, and error_function_c4_aci.
Confidence intervals around the best-fit values can now be calculated automatically by the fitting functions, or manually using three new functions in the package namespace: confidence_intervals_c3_aci, confidence_intervals_c3_variable_j, and confidence_intervals_c4_aci.
The initial guess functions are now created internally instead of being an input argument.
There is a new system for supplying fit options (upper and lower bounds, and which parameters to fit):
Users now only need to specify changes from the default settings.
The order of parameters no longer matters because lists of named elements are used.
Parameters that could be fit, but are not being fit, can either be set to fixed values or to values from a column of an exdf object.
Unknown parameters are now provided in alphabetical order when applicable (such as the first input arguments to calculate_c3_assimilation).
The functions are more tolerant to curves with severe problems (such as negative Ci) that prevent a good fit from being found; rather than throwing an error, the fit functions now silently return NA for all results, along with a message explaining the issue.
Added a function for estimating Rd with the Laisk method: calculate_rd_laisk
A “unit dictionary” was added for internal use; this may be expanded and used more often in the future.
Renamed several variables and input arguments:
Licor files contain a column called alpha, and several different “alphas” were used throughout PhotoGEA. To avoid confusion, the values in PhotoGEA were renamed as follows:
alpha_g: used in C3 assimilation calculations
alpha_pr: used in Gamma_star calculations
alpha_psii: used in C4 assimilation calculations
The acronym “TPU” was used to refer to a process (triose phosphate utilization) and the maximum rate of that process. To avoid confusion, the rate parameter was renamed to Tp
Tests were added for several functions:
calculate_c3_assimilation
fit_c3_aci
fit_c3_variable_j
fit_c4_aci
calculate_c3_limitations_grassi
calculate_c3_limitations_warren
The read_gasex_file function now automatically includes the filename as a column in the resulting exdf object; this helps with troubleshooting problematic curves or files.
Reorganized the variable J fitting functions to be more like fit_c3_aci:
Added calculate_c3_variable_j, initial_guess_c3_variable_j, and fit_c3_variable_j.
Removed dpmn_error_jrv, dpmn_error_jrvtt, dpmn_error_jrv_tau, dpmn_error_jrv_tpu, and photosynthesis_TRF.
Added a new optimizer from the DEoptim package (called optimizer_deoptim) and renamed default_optimizer to optimizer_nmkb. The new optimizer_deoptim is used as the default optimizer for variable J fitting.
Added two new functions for calculating the relative limiting factors of C3 photosynthesis: calculate_c3_limitations_grassi and calculate_c3_limitations_warren.
Added a new function for estimating the operating point from a measured A-Ci curve: estimate_operating_point.
Added two new color specifications (multi_curve_line_colors and multi_curve_point_colors) and used them in vignette examples.
fit_c3_aci and fit_c4_aci now use estimate_operating_point to automatically estimate the operating point and include it with the other fit parameters.
The C3 and C4 A-Ci vignettes now show include the operating point in one of the figures.
Renamed calculate_iwue to calculate_wue and provided documentation for this function, which now calculates two measures of leaf-level water use efficiency.
Provided documentation for factorize_id_column and converted it to an S3 method so it can be applied to data frames and exdf objects.
Removed the process_id_column function since it can easily be replicated using paste.
Stopped reversing colors in xyplot_avg_rc.
Fixed a typo in calculate_c3_assimilation where Rd (the value of day respiration at 25 degrees C) was used in place of Rd_tl (the value of day respiration at the leaf temperature) when calculating net assimilation rates.
Added some developer documentation, an R CMD check GitHub workflow, and a code coverage GitHub workflow.
Added a new convenience function for printing plot objects: pdf_print
Added and/or documented several functions related to isotope discrimination measurements:
calculate_gamma_star
calculate_gm_busch
calculate_gm_ubierna
calculate_isotope_discrimination
calculate_leakiness_ubierna
calculate_ternary_correction
get_oxygen_from_preamble
get_sample_valve_from_filename
pair_gasex_and_tdl
Added two new example data files (licor_for_gm_site11.xslx and tdl_for_gm.dat) for use in examples and vignettes.
Renamed some variables related to isotope measurements; the new terms are more consistent with the way these quantities are typically described:
total_isotope_ratio was renamed to delta_13C
total_mixing_ratio was renamed to total_CO2
Removed a function that was specific to one user (batch_get_genotype_info_from_licor_filename).
Added a new R file (constants.R) to store the values of some constants that appear in multiple functions; this will help ensure that consistent values are used in each instance.
Improved some error handling in read_gasex_file and attempted to clarify instructions for selecting files to load.
Added two basic tests of read_gasex_file using the testthat package.
Modified identify_tdl_cycles and process_tdl_cycle_polynomial so they can handle cycles where there are multiple measurement logs from each valve.
Added several functions for calculating quadratic roots; these are not exported in the package namespace, but are now used in calculate_c3_assimilation and calculate_c4_assimilation to make the code cleaner and more robust.
Other updates to calculate_c3_assimilation:
Choose a minimal RuBP carboxylation rate rather than a minimal net CO2 assimilation rate.
Allow the user to specify a value of alpha (previously this value was hard-coded to 0).
Allow the user to specify two separate curvature values when including co-limitation.
Renamed min_aj_cutoff and max_aj_cutoff to cj_crossover_min and cj_crossover_max to better reflect their purpose.
Added new outputs from fit_c3_aci and fit_c4_aci that include the average values of leaf-temperature-dependent parameters like Vcmax and Rd.
Added options for specifying a flat temperature response when fitting C3 or C4 A-Ci curves; these are available via the two new data sets c3_arrhenius_no_temp and c4_arrhenius_no_temp.
plaintext_licor_file represents a plaintext Licor LI-6800 log file.
c4_aci_1.xlsx and c4_aci_2.xlsx contain examples of C4 A-Ci curves measured with Licor LI-6800 instruments.
Made several significant changes to functions that read data from log files:
Added the ability to read plaintext Licor LI-6800 log files.
Consolidated all file-reading functions into a single one called read_gasex_file.
Depending on user-supplied inputs that specify the instrument and file type, read_gasex_file internally calls either read_licor_6800_Excel, read_licor_6800_plaintext, or read_CR3000 to actually read the data from the file.
When loading LI-6800 Excel log files, it is no longer necessary to provide information about the preamble and data rows.
The previous reading functions read_licor_file and read_tdl_file are now deprecated. Since these functions are commonly used in scripts, a message will be sent to any users who attempt to call these functions. Eventually they will be completely removed from the package namespace.
Made significant changes to the way pressure values are handled:
Added a new function that calculates the total pressure inside a Licor chamber from the separate values of ambient pressure and the chamber overpressure: calculate_total_pressure.
Modified several other functions to just expect a value of total pressure rather than separate values of ambient pressure and chamber overpressure, since this system for storing pressure values is just an artifact of Licor log files:
apply_gm
calculate_c3_assimilation
calculate_gas_properties
fit_c3_aci
Renamed or modified several other functions:
calculate_cc is now apply_gm so it can be used for C3 and C4 photosynthesis calculations.
apply_gm was modified to include partial pressures in its outputs.
fit_c4_aci was modified to make it more like fit_c3_aci.
Added a new input argument to xyplot_avg_rc so that y-axis error bars can be disabled.
Added a new input argument to apply_gm so that drawdown calculations are now optional.
Improved some of the error messages generated by check_licor_data.
Included ATP and NADPH usage in the C3 assimilation calculations.
Added several other new functions:
initial_guess_c4_aci makes initial guesses for C4 photosynthetic parameters.
calculate_c4_assimilation calculates assimilation values using S. von Caemmerer’s model for C4 photosynthesis.
Added a new vignette demonstrating how to analyze C4 A-Ci curve data.
Modified fit_c3_aci so it now ensures that the initial guess lies within (and not on) the bounds. This is a requirement for the dfoptim::nmkb solver and presumably other bounded optimizers as well.
Moved the dfoptim package from Suggests to Imports because it is used by an essential part of PhotoGEA.
The check_required_variables function was moved to the package namespace.
Added and/or documented several functions:
A default optimizer function (default_optimizer).
A function for guessing for C3 parameter values (initial_guess_c3_aci).
A function for identifying columns that take a single unique value (identifier_columns).
A function that calculates error metrics from the values of residuals, including the root mean squared error (RMSE) and several others (residual_stats).
Small changes were made to several functions:
cbind.exdf no longer requires all objects to have the same number of rows.
[<-.exdf now allows users to remove columns by setting their values to NULL.
The conversion of the timestamp column to POSIXlt in read_tdl_file and read_licor_file can now be skipped by setting timestamp_colname to NA.
Significant changes have been made to fit_c3_aci:
It now has an option to set a curvature value that can be used to allow co-limitation of the net assimilation rate.
It now uses initial_guess_c3_aci to generate initial guesses by default.
It now has an option to fix certain parameter values (such as TPU) and exclude them from the fitting process via a new input argument called fixed.
Its output now includes residuals and error metrics such as the RMSE.
fit_ball_berry also returns residuals and error metrics in its output.
Two new vignettes were added:
“Creating Your Own Processing Tools” discusses the best practices for writing custom functions for processing. As an example, a function for fitting a rectangular hyperbola to an A-Ci curve is developed.
“Combining PhotoGEA With Other Packages” discusses best practices for writing wrappers for processing tools from other packages. As an example, a wrapper for plantecophys::fitaci is developed.
Several other vignettes were modified:
The “Analyzing C3 A-Ci Curves” vignette has been updated to include the new features of fit_c3_aci.
The “Developing a Data Analysis Pipeline” vignette has been updated to include the new functions added in this version.
The “Working With Extended Data Frames” vignette has been updated to include some new diagrams.
The “Getting Started With PhotoGEA” vignette has been updated to include links to the new vignettes.
Added a new vignette demonstrating how to analyze C3 A-Ci curve data.
Included TPU in the output from calculate_c3_assimilation.
In the organize_response_curve_data function, changed the default value of the ordering_column_tolerance column to Inf to disable this check by default, since we often want to reorder using a column like Ci that does not follow the same sequence of values in every curve.
Added new input arguments to fit_c3_aci: min_aj_cutoff and max_aj_cutoff, which provide a way to constrain the range of Cc where Aj is allowed to be the limiting assimilation rate.
Added new function for calibrating TDL data (process_tdl_cycle_polynomial) and an option for using it in the gm_from_tdl script.
Modified Licor-TDL pairing to stop assuming a particular relationship between the sample and reference valve numbers.
Fixed an issue with calculate_c3_assimilation that was causing it to report incorrect An values at low Cc.
Fixed several typos where Ac was misidentified as the “RuBP-limited” rate; in fact, it is the RuBP-saturated rate but is more commonly referred to as the rubisco-limited rate.
This version adds a substantial amount of documentation, including a vignette describing how to use PhotoGEA to analyze TDL data.
While working on documentation, several changes were made to key functions so they behaved more reasonably and became easier to document.
Changes to functions and their documentation includes:
exdf objects can now be initialized from just a data frame; in this case, units and categories will all be NA.
apply_fit_across_reps and process_tdl_cycles have been replaced by two new lower-level functions that offer more generality: by.exdf and consolidate.
check_response_curve_data and check_signal_averaging_data have been consolidated into one function: check_licor_data.
An exdf method for cbind has been added.
exclude_tdl_cycles and extract_tdl_valve have been removed since they can easily be reproduced with more basic exdf functionality.
All functions for creating or modifying Excel files have been removed since we no longer want to follow this strategy.
basic_stats and the “Variable J” example script have been fixed.
Changes in PhotoGEA version 0.2.0 (2022-07-26)
This version adds a substantial amount of documentation, including a vignette describing how to use PhotoGEA to analyze Ball-Berry data.
While working on documentation, several changes were made to key functions so they behaved more reasonably and became easier to document.
At the moment, basic_stats and the “Variable J” example script have been broken.
Changes in PhotoGEA version 0.1.0
This is the first version of PhotoGEA. At this point, the package is in a state of rapid development, and not all changes will be described here.
We are reserving version 1.0.0 for the first version where all R package functions and data sets have been documented; until then, major changes should only increase the minor version number.
